API

Top level user functions:

all(a[, axis, out, keepdims]) Test whether all array elements along a given axis evaluate to True.
angle(x[, deg]) Return the angle of the complex argument.
any(a[, axis, out, keepdims]) Test whether any array element along a given axis evaluates to True.
arange(*args, **kwargs) Return evenly spaced values from start to stop with step size step.
arccos(x[, out]) Trigonometric inverse cosine, element-wise.
arccosh(x[, out]) Inverse hyperbolic cosine, element-wise.
arcsin(x[, out]) Inverse sine, element-wise.
arcsinh(x[, out]) Inverse hyperbolic sine element-wise.
arctan(x[, out]) Trigonometric inverse tangent, element-wise.
arctan2(x1, x2[, out]) Element-wise arc tangent of x1/x2 choosing the quadrant correctly.
arctanh(x[, out]) Inverse hyperbolic tangent element-wise.
argmax(a[, axis, out]) Returns the indices of the maximum values along an axis.
argmin(a[, axis, out]) Returns the indices of the minimum values along an axis.
around(a[, decimals, out]) Evenly round to the given number of decimals.
array(object[, dtype, copy, order, subok, ndmin]) Create an array.
bincount(x[, weights, minlength]) Count number of occurrences of each value in array of non-negative ints.
broadcast_to(array, shape[, subok]) Broadcast an array to a new shape.
coarsen(reduction, x, axes[, trim_excess]) Coarsen array by applying reduction to fixed size neighborhoods
ceil(x[, out]) Return the ceiling of the input, element-wise.
choose(a, choices[, out, mode]) Construct an array from an index array and a set of arrays to choose from.
clip(*args, **kwargs) Clip (limit) the values in an array.
compress(condition, a[, axis, out]) Return selected slices of an array along given axis.
concatenate(seq[, axis, ...]) Concatenate arrays along an existing axis
conj(x[, out]) Return the complex conjugate, element-wise.
copysign(x1, x2[, out]) Change the sign of x1 to that of x2, element-wise.
corrcoef(x[, y, rowvar, bias, ddof]) Return Pearson product-moment correlation coefficients.
cos(x[, out]) Cosine element-wise.
cosh(x[, out]) Hyperbolic cosine, element-wise.
cov(m[, y, rowvar, bias, ddof, fweights, ...]) Estimate a covariance matrix, given data and weights.
cumprod(a[, axis, dtype, out]) Return the cumulative product of elements along a given axis.
cumsum(a[, axis, dtype, out]) Return the cumulative sum of the elements along a given axis.
deg2rad(x[, out]) Convert angles from degrees to radians.
degrees(x[, out]) Convert angles from radians to degrees.
diag(v[, k]) Extract a diagonal or construct a diagonal array.
digitize(x, bins[, right]) Return the indices of the bins to which each value in input array belongs.
dot(a, b[, out]) Dot product of two arrays.
dstack(tup) Stack arrays in sequence depth wise (along third axis).
empty Blocked variant of empty
exp(x[, out]) Calculate the exponential of all elements in the input array.
expm1(x[, out]) Calculate exp(x) - 1 for all elements in the array.
eye(N, chunks[, M, k, dtype]) Return a 2-D Array with ones on the diagonal and zeros elsewhere.
fabs(x[, out]) Compute the absolute values element-wise.
fix(*args, **kwargs) Round to nearest integer towards zero.
floor(x[, out]) Return the floor of the input, element-wise.
fmax(x1, x2[, out]) Element-wise maximum of array elements.
fmin(x1, x2[, out]) Element-wise minimum of array elements.
fmod(x1, x2[, out]) Return the element-wise remainder of division.
frexp(x[, out1, out2]) Decompose the elements of x into mantissa and twos exponent.
fromfunction(function, shape, **kwargs) Construct an array by executing a function over each coordinate.
full Blocked variant of full
histogram(a[, bins, range, normed, weights, ...]) Blocked variant of numpy.histogram.
hstack(tup) Stack arrays in sequence horizontally (column wise).
hypot(x1, x2[, out]) Given the “legs” of a right triangle, return its hypotenuse.
imag(*args, **kwargs) Return the imaginary part of the elements of the array.
indices(dimensions[, dtype, chunks]) Implements NumPy’s indices for Dask Arrays.
insert(arr, obj, values[, axis]) Insert values along the given axis before the given indices.
isclose(a, b[, rtol, atol, equal_nan]) Returns a boolean array where two arrays are element-wise equal within a tolerance.
iscomplex(*args, **kwargs) Returns a bool array, where True if input element is complex.
isfinite(x[, out]) Test element-wise for finiteness (not infinity or not Not a Number).
isinf(x[, out]) Test element-wise for positive or negative infinity.
isnan(x[, out]) Test element-wise for NaN and return result as a boolean array.
isnull(values) pandas.isnull for dask arrays
isreal(*args, **kwargs) Returns a bool array, where True if input element is real.
ldexp(x1, x2[, out]) Returns x1 * 2**x2, element-wise.
linspace(start, stop[, num, chunks, dtype]) Return num evenly spaced values over the closed interval [start, stop].
log(x[, out]) Natural logarithm, element-wise.
log10(x[, out]) Return the base 10 logarithm of the input array, element-wise.
log1p(x[, out]) Return the natural logarithm of one plus the input array, element-wise.
log2(x[, out]) Base-2 logarithm of x.
logaddexp(x1, x2[, out]) Logarithm of the sum of exponentiations of the inputs.
logaddexp2(x1, x2[, out]) Logarithm of the sum of exponentiations of the inputs in base-2.
logical_and(x1, x2[, out]) Compute the truth value of x1 AND x2 element-wise.
logical_not(x[, out]) Compute the truth value of NOT x element-wise.
logical_or(x1, x2[, out]) Compute the truth value of x1 OR x2 element-wise.
logical_xor(x1, x2[, out]) Compute the truth value of x1 XOR x2, element-wise.
max(a[, axis, out, keepdims]) Return the maximum of an array or maximum along an axis.
maximum(x1, x2[, out]) Element-wise maximum of array elements.
mean(a[, axis, dtype, out, keepdims]) Compute the arithmetic mean along the specified axis.
min(a[, axis, out, keepdims]) Return the minimum of an array or minimum along an axis.
minimum(x1, x2[, out]) Element-wise minimum of array elements.
modf(x[, out1, out2]) Return the fractional and integral parts of an array, element-wise.
moment(a, order[, axis, dtype, keepdims, ...])
nanargmax(x, axis, **kwargs)
nanargmin(x, axis, **kwargs)
nancumprod(a[, axis, dtype, out]) Return the cumulative product of array elements over a given axis treating Not a Numbers (NaNs) as one.
nancumsum(a[, axis, dtype, out]) Return the cumulative sum of array elements over a given axis treating Not a Numbers (NaNs) as zero.
nanmax(a[, axis, out, keepdims]) Return the maximum of an array or maximum along an axis, ignoring any NaNs.
nanmean(a[, axis, dtype, out, keepdims]) Compute the arithmetic mean along the specified axis, ignoring NaNs.
nanmin(a[, axis, out, keepdims]) Return minimum of an array or minimum along an axis, ignoring any NaNs.
nanprod(a[, axis, dtype, out, keepdims]) Return the product of array elements over a given axis treating Not a Numbers (NaNs) as ones.
nanstd(a[, axis, dtype, out, ddof, keepdims]) Compute the standard deviation along the specified axis, while ignoring NaNs.
nansum(a[, axis, dtype, out, keepdims]) Return the sum of array elements over a given axis treating Not a Numbers (NaNs) as zero.
nanvar(a[, axis, dtype, out, ddof, keepdims]) Compute the variance along the specified axis, while ignoring NaNs.
nextafter(x1, x2[, out]) Return the next floating-point value after x1 towards x2, element-wise.
notnull(values) pandas.notnull for dask arrays
ones Blocked variant of ones
percentile(a, q[, interpolation]) Approximate percentile of 1-D array
prod(a[, axis, dtype, out, keepdims]) Return the product of array elements over a given axis.
rad2deg(x[, out]) Convert angles from radians to degrees.
radians(x[, out]) Convert angles from degrees to radians.
ravel(a[, order]) Return a contiguous flattened array.
real(*args, **kwargs) Return the real part of the elements of the array.
rechunk(x, chunks[, threshold, block_size_limit]) Convert blocks in dask array x for new chunks.
repeat(a, repeats[, axis]) Repeat elements of an array.
reshape(x, shape) Reshape array to new shape
rint(x[, out]) Round elements of the array to the nearest integer.
roll(a, shift[, axis]) Roll array elements along a given axis.
round(a[, decimals, out]) Round an array to the given number of decimals.
sign(x[, out]) Returns an element-wise indication of the sign of a number.
signbit(x[, out]) Returns element-wise True where signbit is set (less than zero).
sin(x[, out]) Trigonometric sine, element-wise.
sinh(x[, out]) Hyperbolic sine, element-wise.
sqrt(x[, out]) Return the positive square-root of an array, element-wise.
square(x[, out]) Return the element-wise square of the input.
squeeze(a[, axis]) Remove single-dimensional entries from the shape of an array.
stack(seq[, axis]) Stack arrays along a new axis
std(a[, axis, dtype, out, ddof, keepdims]) Compute the standard deviation along the specified axis.
sum(a[, axis, dtype, out, keepdims]) Sum of array elements over a given axis.
take(a, indices[, axis, out, mode]) Take elements from an array along an axis.
tan(x[, out]) Compute tangent element-wise.
tanh(x[, out]) Compute hyperbolic tangent element-wise.
tensordot(a, b[, axes]) Compute tensor dot product along specified axes for arrays >= 1-D.
tile(A, reps) Construct an array by repeating A the number of times given by reps.
topk(k, x) The top k elements of an array
transpose(a[, axes]) Permute the dimensions of an array.
tril(m[, k]) Lower triangle of an array with elements above the k-th diagonal zeroed.
triu(m[, k]) Upper triangle of an array with elements above the k-th diagonal zeroed.
trunc(x[, out]) Return the truncated value of the input, element-wise.
unique(ar[, return_index, return_inverse, ...]) Find the unique elements of an array.
var(a[, axis, dtype, out, ddof, keepdims]) Compute the variance along the specified axis.
vnorm(a[, ord, axis, dtype, keepdims, ...]) Vector norm
vstack(tup) Stack arrays in sequence vertically (row wise).
where(condition, [x, y]) Return elements, either from x or y, depending on condition.
zeros Blocked variant of zeros

Fast Fourier Transforms

fft.fft_wrap(fft_func[, kind, dtype]) Wrap 1D complex FFT functions
fft.fft(a[, n, axis]) Wrapping of numpy.fft.fftpack.fft
fft.fft2(a[, s, axes]) Wrapping of numpy.fft.fftpack.fft2
fft.fftn(a[, s, axes]) Wrapping of numpy.fft.fftpack.fftn
fft.ifft(a[, n, axis]) Wrapping of numpy.fft.fftpack.ifft
fft.ifft2(a[, s, axes]) Wrapping of numpy.fft.fftpack.ifft2
fft.ifftn(a[, s, axes]) Wrapping of numpy.fft.fftpack.ifftn
fft.rfft(a[, n, axis]) Wrapping of numpy.fft.fftpack.rfft
fft.rfft2(a[, s, axes]) Wrapping of numpy.fft.fftpack.rfft2
fft.rfftn(a[, s, axes]) Wrapping of numpy.fft.fftpack.rfftn
fft.irfft(a[, n, axis]) Wrapping of numpy.fft.fftpack.irfft
fft.irfft2(a[, s, axes]) Wrapping of numpy.fft.fftpack.irfft2
fft.irfftn(a[, s, axes]) Wrapping of numpy.fft.fftpack.irfftn
fft.hfft(a[, n, axis]) Wrapping of numpy.fft.fftpack.hfft
fft.ihfft(a[, n, axis]) Wrapping of numpy.fft.fftpack.ihfft
fft.fftfreq(n[, d]) Return the Discrete Fourier Transform sample frequencies.
fft.rfftfreq(n[, d]) Return the Discrete Fourier Transform sample frequencies (for usage with rfft, irfft).
fft.fftshift(x[, axes]) Shift the zero-frequency component to the center of the spectrum.
fft.ifftshift(x[, axes]) The inverse of fftshift.

Linear Algebra

linalg.cholesky(a[, lower]) Returns the Cholesky decomposition, \(A = L L^*\) or \(A = U^* U\) of a Hermitian positive-definite matrix A.
linalg.inv(a) Compute the inverse of a matrix with LU decomposition and forward / backward substitutions.
linalg.lstsq(a, b) Return the least-squares solution to a linear matrix equation using QR decomposition.
linalg.lu(a) Compute the lu decomposition of a matrix.
linalg.qr(a[, name]) Compute the qr factorization of a matrix.
linalg.solve(a, b[, sym_pos]) Solve the equation a x = b for x.
linalg.solve_triangular(a, b[, lower]) Solve the equation a x = b for x, assuming a is a triangular matrix.
linalg.svd(a[, name]) Compute the singular value decomposition of a matrix.
linalg.svd_compressed(a, k[, n_power_iter, ...]) Randomly compressed rank-k thin Singular Value Decomposition.
linalg.tsqr(data[, name, compute_svd]) Direct Tall-and-Skinny QR algorithm

Random

random.beta(a, b[, size]) Draw samples from a Beta distribution.
random.binomial(n, p[, size]) Draw samples from a binomial distribution.
random.chisquare(df[, size]) Draw samples from a chi-square distribution.
random.different_seeds
random.exponential([scale, size]) Draw samples from an exponential distribution.
random.f(dfnum, dfden[, size]) Draw samples from an F distribution.
random.gamma(shape[, scale, size]) Draw samples from a Gamma distribution.
random.geometric(p[, size]) Draw samples from the geometric distribution.
random.gumbel([loc, scale, size]) Draw samples from a Gumbel distribution.
random.hypergeometric(ngood, nbad, nsample) Draw samples from a Hypergeometric distribution.
random.laplace([loc, scale, size]) Draw samples from the Laplace or double exponential distribution with specified location (or mean) and scale (decay).
random.logistic([loc, scale, size]) Draw samples from a logistic distribution.
random.lognormal([mean, sigma, size]) Draw samples from a log-normal distribution.
random.logseries(p[, size]) Draw samples from a logarithmic series distribution.
random.negative_binomial(n, p[, size]) Draw samples from a negative binomial distribution.
random.noncentral_chisquare(df, nonc[, size]) Draw samples from a noncentral chi-square distribution.
random.noncentral_f(dfnum, dfden, nonc[, size]) Draw samples from the noncentral F distribution.
random.normal([loc, scale, size]) Draw random samples from a normal (Gaussian) distribution.
random.pareto(a[, size]) Draw samples from a Pareto II or Lomax distribution with specified shape.
random.poisson([lam, size]) Draw samples from a Poisson distribution.
random.power(a[, size]) Draws samples in [0, 1] from a power distribution with positive exponent a - 1.
random.random([size]) Return random floats in the half-open interval [0.0, 1.0).
random.random_sample([size]) Return random floats in the half-open interval [0.0, 1.0).
random.rayleigh([scale, size]) Draw samples from a Rayleigh distribution.
random.standard_cauchy([size]) Draw samples from a standard Cauchy distribution with mode = 0.
random.standard_exponential([size]) Draw samples from the standard exponential distribution.
random.standard_gamma(shape[, size]) Draw samples from a standard Gamma distribution.
random.standard_normal([size]) Draw samples from a standard Normal distribution (mean=0, stdev=1).
random.standard_t(df[, size]) Draw samples from a standard Student’s t distribution with df degrees of freedom.
random.triangular(left, mode, right[, size]) Draw samples from the triangular distribution over the interval [left, right].
random.uniform([low, high, size]) Draw samples from a uniform distribution.
random.vonmises(mu, kappa[, size]) Draw samples from a von Mises distribution.
random.wald(mean, scale[, size]) Draw samples from a Wald, or inverse Gaussian, distribution.
random.weibull(a[, size]) Draw samples from a Weibull distribution.
random.zipf(a[, size]) Standard distributions

Stats

stats.ttest_ind(a, b[, axis, equal_var]) Calculates the T-test for the means of two independent samples of scores.
stats.ttest_1samp(a, popmean[, axis, nan_policy]) Calculates the T-test for the mean of ONE group of scores.
stats.ttest_rel(a, b[, axis, nan_policy]) Calculates the T-test on TWO RELATED samples of scores, a and b.
stats.chisquare(f_obs[, f_exp, ddof, axis]) Calculates a one-way chi square test.
stats.power_divergence(f_obs[, f_exp, ddof, ...]) Cressie-Read power divergence statistic and goodness of fit test.
stats.skew(a[, axis, bias, nan_policy]) Computes the skewness of a data set.
stats.skewtest(a[, axis, nan_policy]) Tests whether the skew is different from the normal distribution.
stats.kurtosis(a[, axis, fisher, bias, ...]) Computes the kurtosis (Fisher or Pearson) of a dataset.
stats.kurtosistest(a[, axis, nan_policy]) Tests whether a dataset has normal kurtosis
stats.normaltest(a[, axis, nan_policy]) Tests whether a sample differs from a normal distribution.
stats.f_oneway(*args) Performs a 1-way ANOVA.
stats.moment(a[, moment, axis, nan_policy]) Calculates the nth moment about the mean for a sample.

Image Support

image.imread(filename[, imread, preprocess]) Read a stack of images into a dask array

Slightly Overlapping Ghost Computations

ghost.ghost(x, depth, boundary) Share boundaries between neighboring blocks
ghost.map_overlap(x, func, depth[, ...])

Create and Store Arrays

from_array(x, chunks[, name, lock, asarray, ...]) Create dask array from something that looks like an array
from_delayed(value, shape, dtype[, name]) Create a dask array from a dask delayed value
from_npy_stack(dirname[, mmap_mode]) Load dask array from stack of npy files
store(sources, targets[, lock, regions, compute]) Store dask arrays in array-like objects, overwrite data in target
to_hdf5(filename, *args, **kwargs) Store arrays in HDF5 file
to_npy_stack(dirname, x[, axis]) Write dask array to a stack of .npy files

Internal functions

map_blocks(func, *args, **kwargs) Map a function across all blocks of a dask array.
atop(func, out_ind, *args, **kwargs) Tensor operation: Generalized inner and outer products
top(func, output, out_indices, ...) Tensor operation

Other functions

dask.array.from_array(x, chunks, name=None, lock=False, asarray=True, fancy=True, getitem=None)

Create dask array from something that looks like an array

Input must have a .shape and support numpy-style slicing.

Parameters:

x : array_like

chunks : int, tuple

How to chunk the array. Must be one of the following forms: - A blocksize like 1000. - A blockshape like (1000, 1000). - Explicit sizes of all blocks along all dimensions

like ((1000, 1000, 500), (400, 400)).

name : str, optional

The key name to use for the array. Defaults to a hash of x. Use name=False to generate a random name instead of hashing (fast)

lock : bool or Lock, optional

If x doesn’t support concurrent reads then provide a lock here, or pass in True to have dask.array create one for you.

asarray : bool, optional

If True (default), then chunks will be converted to instances of ndarray. Set to False to pass passed chunks through unchanged.

fancy : bool, optional

If x doesn’t support fancy indexing (e.g. indexing with lists or arrays) then set to False. Default is True.

Examples

>>> x = h5py.File('...')['/data/path']  
>>> a = da.from_array(x, chunks=(1000, 1000))

If your underlying datastore does not support concurrent reads then include the lock=True keyword argument or lock=mylock if you want multiple arrays to coordinate around the same lock.

>>> a = da.from_array(x, chunks=(1000, 1000), lock=True)
dask.array.from_delayed(value, shape, dtype, name=None)

Create a dask array from a dask delayed value

This routine is useful for constructing dask arrays in an ad-hoc fashion using dask delayed, particularly when combined with stack and concatenate.

The dask array will consist of a single chunk.

Examples

>>> from dask import delayed
>>> value = delayed(np.ones)(5)
>>> array = from_delayed(value, (5,), float)
>>> array
dask.array<from-value, shape=(5,), dtype=float64, chunksize=(5,)>
>>> array.compute()
array([ 1.,  1.,  1.,  1.,  1.])
dask.array.store(sources, targets, lock=True, regions=None, compute=True, **kwargs)

Store dask arrays in array-like objects, overwrite data in target

This stores dask arrays into object that supports numpy-style setitem indexing. It stores values chunk by chunk so that it does not have to fill up memory. For best performance you can align the block size of the storage target with the block size of your array.

If your data fits in memory then you may prefer calling np.array(myarray) instead.

Parameters:

sources: Array or iterable of Arrays

targets: array-like or iterable of array-likes

These should support setitem syntax target[10:20] = ...

lock: boolean or threading.Lock, optional

Whether or not to lock the data stores while storing. Pass True (lock each file individually), False (don’t lock) or a particular threading.Lock object to be shared among all writes.

regions: tuple of slices or iterable of tuple of slices

Each region tuple in regions should be such that target[region].shape = source.shape for the corresponding source and target in sources and targets, respectively.

compute: boolean, optional

If true compute immediately, return dask.delayed.Delayed otherwise

Examples

>>> x = ...  

>>> import h5py  
>>> f = h5py.File('myfile.hdf5')  
>>> dset = f.create_dataset('/data', shape=x.shape,
...                                  chunks=x.chunks,
...                                  dtype='f8')  

>>> store(x, dset)

Alternatively store many arrays at the same time

>>> store([x, y, z], [dset1, dset2, dset3])
dask.array.topk(k, x)

The top k elements of an array

Returns the k greatest elements of the array in sorted order. Only works on arrays of a single dimension.

This assumes that k is small. All results will be returned in a single chunk.

Examples

>>> x = np.array([5, 1, 3, 6])
>>> d = from_array(x, chunks=2)
>>> d.topk(2).compute()
array([6, 5])
dask.array.coarsen(reduction, x, axes, trim_excess=False)

Coarsen array by applying reduction to fixed size neighborhoods

Parameters:

reduction: function

Function like np.sum, np.mean, etc...

x: np.ndarray

Array to be coarsened

axes: dict

Mapping of axis to coarsening factor

Examples

>>> x = np.array([1, 2, 3, 4, 5, 6])
>>> coarsen(np.sum, x, {0: 2})
array([ 3,  7, 11])
>>> coarsen(np.max, x, {0: 3})
array([3, 6])

Provide dictionary of scale per dimension

>>> x = np.arange(24).reshape((4, 6))
>>> x
array([[ 0,  1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17],
       [18, 19, 20, 21, 22, 23]])

>>> coarsen(np.min, x, {0: 2, 1: 3})
array([[ 0,  3],
       [12, 15]])

You must avoid excess elements explicitly

>>> x = np.array([1, 2, 3, 4, 5, 6, 7, 8])
>>> coarsen(np.min, x, {0: 3}, trim_excess=True)
array([1, 4])
dask.array.stack(seq, axis=0)

Stack arrays along a new axis

Given a sequence of dask Arrays form a new dask Array by stacking them along a new dimension (axis=0 by default)

See also

concatenate

Examples

Create slices

>>> import dask.array as da
>>> import numpy as np

>>> data = [from_array(np.ones((4, 4)), chunks=(2, 2))
...          for i in range(3)]

>>> x = da.stack(data, axis=0)
>>> x.shape
(3, 4, 4)

>>> da.stack(data, axis=1).shape
(4, 3, 4)

>>> da.stack(data, axis=-1).shape
(4, 4, 3)

Result is a new dask Array

dask.array.concatenate(seq, axis=0, allow_unknown_chunksizes=False)

Concatenate arrays along an existing axis

Given a sequence of dask Arrays form a new dask Array by stacking them along an existing dimension (axis=0 by default)

Parameters:

seq: list of dask.arrays

axis: int

Dimension along which to align all of the arrays

allow_unknown_chunksizes: bool

Allow unknown chunksizes, such as come from converting from dask dataframes. Dask.array is unable to verify that chunks line up. If data comes from differently aligned sources then this can cause unexpected results.

See also

stack

Examples

Create slices

>>> import dask.array as da
>>> import numpy as np

>>> data = [from_array(np.ones((4, 4)), chunks=(2, 2))
...          for i in range(3)]

>>> x = da.concatenate(data, axis=0)
>>> x.shape
(12, 4)

>>> da.concatenate(data, axis=1).shape
(4, 12)

Result is a new dask Array

dask.array.all(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Test whether all array elements along a given axis evaluate to True.

Parameters:

a : array_like

Input array or object that can be converted to an array.

axis : None or int or tuple of ints, optional

Axis or axes along which a logical AND reduction is performed. The default (axis = None) is to perform a logical AND over all the dimensions of the input array. axis may be negative, in which case it counts from the last to the first axis.

New in version 1.7.0.

If this is a tuple of ints, a reduction is performed on multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output and its type is preserved (e.g., if dtype(out) is float, the result will consist of 0.0’s and 1.0’s). See doc.ufuncs (Section “Output arguments”) for more details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the all method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

all : ndarray, bool

A new boolean or array is returned unless out is specified, in which case a reference to out is returned.

See also

ndarray.all
equivalent method
any
Test whether any element along a given axis evaluates to True.

Notes

Not a Number (NaN), positive infinity and negative infinity evaluate to True because these are not equal to zero.

Examples

>>> np.all([[True,False],[True,True]])
False

>>> np.all([[True,False],[True,True]], axis=0)
array([ True, False], dtype=bool)

>>> np.all([-1, 4, 5])
True

>>> np.all([1.0, np.nan])
True

>>> o=np.array([False])
>>> z=np.all([-1, 4, 5], out=o)
>>> id(z), id(o), z                             
(28293632, 28293632, array([ True], dtype=bool))
dask.array.angle(x, deg=0)

Return the angle of the complex argument.

Parameters:

z : array_like

A complex number or sequence of complex numbers.

deg : bool, optional

Return angle in degrees if True, radians if False (default).
Returns:

angle : ndarray or scalar

The counterclockwise angle from the positive real axis on the complex plane, with dtype as numpy.float64.

See also

arctan2, absolute

Examples

>>> np.angle([1.0, 1.0j, 1+1j])               # in radians  
array([ 0.        ,  1.57079633,  0.78539816])
>>> np.angle(1+1j, deg=True)                  # in degrees  
45.0
dask.array.any(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Test whether any array element along a given axis evaluates to True.

Returns single boolean unless axis is not None

Parameters:

a : array_like

Input array or object that can be converted to an array.

axis : None or int or tuple of ints, optional

Axis or axes along which a logical OR reduction is performed. The default (axis = None) is to perform a logical OR over all the dimensions of the input array. axis may be negative, in which case it counts from the last to the first axis.

New in version 1.7.0.

If this is a tuple of ints, a reduction is performed on multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output and its type is preserved (e.g., if it is of type float, then it will remain so, returning 1.0 for True and 0.0 for False, regardless of the type of a). See doc.ufuncs (Section “Output arguments”) for details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the any method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

any : bool or ndarray

A new boolean or ndarray is returned unless out is specified, in which case a reference to out is returned.

See also

ndarray.any
equivalent method
all
Test whether all elements along a given axis evaluate to True.

Notes

Not a Number (NaN), positive infinity and negative infinity evaluate to True because these are not equal to zero.

Examples

>>> np.any([[True, False], [True, True]])
True

>>> np.any([[True, False], [False, False]], axis=0)
array([ True, False], dtype=bool)

>>> np.any([-1, 0, 5])
True

>>> np.any(np.nan)
True

>>> o=np.array([False])
>>> z=np.any([-1, 4, 5], out=o)
>>> z, o
(array([ True], dtype=bool), array([ True], dtype=bool))
>>> # Check now that z is a reference to o
>>> z is o
True
>>> id(z), id(o) # identity of z and o              
(191614240, 191614240)
dask.array.arange(*args, **kwargs)

Return evenly spaced values from start to stop with step size step.

The values are half-open [start, stop), so including start and excluding stop. This is basically the same as python’s range function but for dask arrays.

When using a non-integer step, such as 0.1, the results will often not be consistent. It is better to use linspace for these cases.

Parameters:

start : int, optional

The starting value of the sequence. The default is 0.

stop : int

The end of the interval, this value is excluded from the interval.

step : int, optional

The spacing between the values. The default is 1 when not specified. The last value of the sequence.

chunks : int

The number of samples on each block. Note that the last block will have fewer samples if len(array) % chunks != 0.
Returns:

samples : dask array

See also

dask.array.linspace

dask.array.arccos(x[, out])

Trigonometric inverse cosine, element-wise.

The inverse of cos so that, if y = cos(x), then x = arccos(y).

Parameters:

x : array_like

x-coordinate on the unit circle. For real arguments, the domain is [-1, 1].

out : ndarray, optional

Array of the same shape as a, to store results in. See doc.ufuncs (Section “Output arguments”) for more details.
Returns:

angle : ndarray

The angle of the ray intersecting the unit circle at the given x-coordinate in radians [0, pi]. If x is a scalar then a scalar is returned, otherwise an array of the same shape as x is returned.

See also

cos, arctan, arcsin, emath.arccos

Notes

arccos is a multivalued function: for each x there are infinitely many numbers z such that cos(z) = x. The convention is to return the angle z whose real part lies in [0, pi].

For real-valued input data types, arccos always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, arccos is a complex analytic function that has branch cuts [-inf, -1] and [1, inf] and is continuous from above on the former and from below on the latter.

The inverse cos is also known as acos or cos^-1.

References

M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 79. http://www.math.sfu.ca/~cbm/aands/

Examples

We expect the arccos of 1 to be 0, and of -1 to be pi:

>>> np.arccos([1, -1])  
array([ 0.        ,  3.14159265])

Plot arccos:

>>> import matplotlib.pyplot as plt  
>>> x = np.linspace(-1, 1, num=100)  
>>> plt.plot(x, np.arccos(x))  
>>> plt.axis('tight')  
>>> plt.show()
dask.array.arccosh(x[, out])

Inverse hyperbolic cosine, element-wise.

Parameters:

x : array_like

Input array.

out : ndarray, optional

Array of the same shape as x, to store results in. See doc.ufuncs (Section “Output arguments”) for details.
Returns:

arccosh : ndarray

Array of the same shape as x.

See also

cosh, arcsinh, sinh, arctanh, tanh

Notes

arccosh is a multivalued function: for each x there are infinitely many numbers z such that cosh(z) = x. The convention is to return the z whose imaginary part lies in [-pi, pi] and the real part in [0, inf].

For real-valued input data types, arccosh always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, arccosh is a complex analytical function that has a branch cut [-inf, 1] and is continuous from above on it.

References

[R207208]M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 86. http://www.math.sfu.ca/~cbm/aands/
[R208208]Wikipedia, “Inverse hyperbolic function”, http://en.wikipedia.org/wiki/Arccosh

Examples

>>> np.arccosh([np.e, 10.0])  
array([ 1.65745445,  2.99322285])
>>> np.arccosh(1)  
0.0
dask.array.arcsin(x[, out])

Inverse sine, element-wise.

Parameters:

x : array_like

y-coordinate on the unit circle.

out : ndarray, optional

Array of the same shape as x, in which to store the results. See doc.ufuncs (Section “Output arguments”) for more details.
Returns:

angle : ndarray

The inverse sine of each element in x, in radians and in the closed interval [-pi/2, pi/2]. If x is a scalar, a scalar is returned, otherwise an array.

See also

sin, cos, arccos, tan, arctan, arctan2, emath.arcsin

Notes

arcsin is a multivalued function: for each x there are infinitely many numbers z such that \(sin(z) = x\). The convention is to return the angle z whose real part lies in [-pi/2, pi/2].

For real-valued input data types, arcsin always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, arcsin is a complex analytic function that has, by convention, the branch cuts [-inf, -1] and [1, inf] and is continuous from above on the former and from below on the latter.

The inverse sine is also known as asin or sin^{-1}.

References

Abramowitz, M. and Stegun, I. A., Handbook of Mathematical Functions, 10th printing, New York: Dover, 1964, pp. 79ff. http://www.math.sfu.ca/~cbm/aands/

Examples

>>> np.arcsin(1)     # pi/2  
1.5707963267948966
>>> np.arcsin(-1)    # -pi/2  
-1.5707963267948966
>>> np.arcsin(0)  
0.0
dask.array.arcsinh(x[, out])

Inverse hyperbolic sine element-wise.

Parameters:

x : array_like

Input array.

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

out : ndarray

Array of of the same shape as x.

Notes

arcsinh is a multivalued function: for each x there are infinitely many numbers z such that sinh(z) = x. The convention is to return the z whose imaginary part lies in [-pi/2, pi/2].

For real-valued input data types, arcsinh always returns real output. For each value that cannot be expressed as a real number or infinity, it returns nan and sets the invalid floating point error flag.

For complex-valued input, arccos is a complex analytical function that has branch cuts [1j, infj] and [-1j, -infj] and is continuous from the right on the former and from the left on the latter.

The inverse hyperbolic sine is also known as asinh or sinh^-1.

References

[R211212]M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 86. http://www.math.sfu.ca/~cbm/aands/
[R212212]Wikipedia, “Inverse hyperbolic function”, http://en.wikipedia.org/wiki/Arcsinh

Examples

>>> np.arcsinh(np.array([np.e, 10.0]))  
array([ 1.72538256,  2.99822295])
dask.array.arctan(x[, out])

Trigonometric inverse tangent, element-wise.

The inverse of tan, so that if y = tan(x) then x = arctan(y).

Parameters:

x : array_like

Input values. arctan is applied to each element of x.
Returns:

out : ndarray

Out has the same shape as x. Its real part is in [-pi/2, pi/2] (arctan(+/-inf) returns +/-pi/2). It is a scalar if x is a scalar.

See also

arctan2
The “four quadrant” arctan of the angle formed by (x, y) and the positive x-axis.
angle
Argument of complex values.

Notes

arctan is a multi-valued function: for each x there are infinitely many numbers z such that tan(z) = x. The convention is to return the angle z whose real part lies in [-pi/2, pi/2].

For real-valued input data types, arctan always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, arctan is a complex analytic function that has [1j, infj] and [-1j, -infj] as branch cuts, and is continuous from the left on the former and from the right on the latter.

The inverse tangent is also known as atan or tan^{-1}.

References

Abramowitz, M. and Stegun, I. A., Handbook of Mathematical Functions, 10th printing, New York: Dover, 1964, pp. 79. http://www.math.sfu.ca/~cbm/aands/

Examples

We expect the arctan of 0 to be 0, and of 1 to be pi/4:

>>> np.arctan([0, 1])  
array([ 0.        ,  0.78539816])

>>> np.pi/4  
0.78539816339744828

Plot arctan:

>>> import matplotlib.pyplot as plt  
>>> x = np.linspace(-10, 10)  
>>> plt.plot(x, np.arctan(x))  
>>> plt.axis('tight')  
>>> plt.show()
dask.array.arctan2(x1, x2[, out])

Element-wise arc tangent of x1/x2 choosing the quadrant correctly.

The quadrant (i.e., branch) is chosen so that arctan2(x1, x2) is the signed angle in radians between the ray ending at the origin and passing through the point (1,0), and the ray ending at the origin and passing through the point (x2, x1). (Note the role reversal: the “y-coordinate” is the first function parameter, the “x-coordinate” is the second.) By IEEE convention, this function is defined for x2 = +/-0 and for either or both of x1 and x2 = +/-inf (see Notes for specific values).

This function is not defined for complex-valued arguments; for the so-called argument of complex values, use angle.

Parameters:

x1 : array_like, real-valued

y-coordinates.

x2 : array_like, real-valued

x-coordinates. x2 must be broadcastable to match the shape of x1 or vice versa.
Returns:

angle : ndarray

Array of angles in radians, in the range [-pi, pi].

See also

arctan, tan, angle

Notes

arctan2 is identical to the atan2 function of the underlying C library. The following special values are defined in the C standard: [R215215]

x1 x2 arctan2(x1,x2)
+/- 0 +0 +/- 0
+/- 0 -0 +/- pi
> 0 +/-inf +0 / +pi
< 0 +/-inf -0 / -pi
+/-inf +inf +/- (pi/4)
+/-inf -inf +/- (3*pi/4)

Note that +0 and -0 are distinct floating point numbers, as are +inf and -inf.

References

[R215215](1, 2) ISO/IEC standard 9899:1999, “Programming language C.”

Examples

Consider four points in different quadrants:

>>> x = np.array([-1, +1, +1, -1])  
>>> y = np.array([-1, -1, +1, +1])  
>>> np.arctan2(y, x) * 180 / np.pi  
array([-135.,  -45.,   45.,  135.])

Note the order of the parameters. arctan2 is defined also when x2 = 0 and at several other special points, obtaining values in the range [-pi, pi]:

>>> np.arctan2([1., -1.], [0., 0.])  
array([ 1.57079633, -1.57079633])
>>> np.arctan2([0., 0., np.inf], [+0., -0., np.inf])  
array([ 0.        ,  3.14159265,  0.78539816])
dask.array.arctanh(x[, out])

Inverse hyperbolic tangent element-wise.

Parameters:

x : array_like

Input array.
Returns:

out : ndarray

Array of the same shape as x.

See also

emath.arctanh

Notes

arctanh is a multivalued function: for each x there are infinitely many numbers z such that tanh(z) = x. The convention is to return the z whose imaginary part lies in [-pi/2, pi/2].

For real-valued input data types, arctanh always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, arctanh is a complex analytical function that has branch cuts [-1, -inf] and [1, inf] and is continuous from above on the former and from below on the latter.

The inverse hyperbolic tangent is also known as atanh or tanh^-1.

References

[R217218]M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 86. http://www.math.sfu.ca/~cbm/aands/
[R218218]Wikipedia, “Inverse hyperbolic function”, http://en.wikipedia.org/wiki/Arctanh

Examples

>>> np.arctanh([0, -0.5])  
array([ 0.        , -0.54930614])
dask.array.argmax(a, axis=None, out=None)

Returns the indices of the maximum values along an axis.

Parameters:

a : array_like

Input array.

axis : int, optional

By default, the index is into the flattened array, otherwise along the specified axis.

out : array, optional

If provided, the result will be inserted into this array. It should be of the appropriate shape and dtype.
Returns:

index_array : ndarray of ints

Array of indices into the array. It has the same shape as a.shape with the dimension along axis removed.

See also

ndarray.argmax, argmin

amax
The maximum value along a given axis.
unravel_index
Convert a flat index into an index tuple.

Notes

In case of multiple occurrences of the maximum values, the indices corresponding to the first occurrence are returned.

Examples

>>> a = np.arange(6).reshape(2,3)
>>> a
array([[0, 1, 2],
       [3, 4, 5]])
>>> np.argmax(a)
5
>>> np.argmax(a, axis=0)
array([1, 1, 1])
>>> np.argmax(a, axis=1)
array([2, 2])

>>> b = np.arange(6)
>>> b[1] = 5
>>> b
array([0, 5, 2, 3, 4, 5])
>>> np.argmax(b) # Only the first occurrence is returned.
1
dask.array.argmin(a, axis=None, out=None)

Returns the indices of the minimum values along an axis.

Parameters:

a : array_like

Input array.

axis : int, optional

By default, the index is into the flattened array, otherwise along the specified axis.

out : array, optional

If provided, the result will be inserted into this array. It should be of the appropriate shape and dtype.
Returns:

index_array : ndarray of ints

Array of indices into the array. It has the same shape as a.shape with the dimension along axis removed.

See also

ndarray.argmin, argmax

amin
The minimum value along a given axis.
unravel_index
Convert a flat index into an index tuple.

Notes

In case of multiple occurrences of the minimum values, the indices corresponding to the first occurrence are returned.

Examples

>>> a = np.arange(6).reshape(2,3)
>>> a
array([[0, 1, 2],
       [3, 4, 5]])
>>> np.argmin(a)
0
>>> np.argmin(a, axis=0)
array([0, 0, 0])
>>> np.argmin(a, axis=1)
array([0, 0])

>>> b = np.arange(6)
>>> b[4] = 0
>>> b
array([0, 1, 2, 3, 0, 5])
>>> np.argmin(b) # Only the first occurrence is returned.
0
dask.array.around(a, decimals=0, out=None)

Evenly round to the given number of decimals.

Parameters:

a : array_like

Input data.

decimals : int, optional

Number of decimal places to round to (default: 0). If decimals is negative, it specifies the number of positions to the left of the decimal point.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary. See doc.ufuncs (Section “Output arguments”) for details.
Returns:

rounded_array : ndarray

An array of the same type as a, containing the rounded values. Unless out was specified, a new array is created. A reference to the result is returned.

The real and imaginary parts of complex numbers are rounded separately. The result of rounding a float is a float.

See also

ndarray.round
equivalent method

ceil, fix, floor, rint, trunc

Notes

For values exactly halfway between rounded decimal values, NumPy rounds to the nearest even value. Thus 1.5 and 2.5 round to 2.0, -0.5 and 0.5 round to 0.0, etc. Results may also be surprising due to the inexact representation of decimal fractions in the IEEE floating point standard [R221222] and errors introduced when scaling by powers of ten.

References

[R221222](1, 2) “Lecture Notes on the Status of IEEE 754”, William Kahan, http://www.cs.berkeley.edu/~wkahan/ieee754status/IEEE754.PDF
[R222222]“How Futile are Mindless Assessments of Roundoff in Floating-Point Computation?”, William Kahan, http://www.cs.berkeley.edu/~wkahan/Mindless.pdf

Examples

>>> np.around([0.37, 1.64])
array([ 0.,  2.])
>>> np.around([0.37, 1.64], decimals=1)
array([ 0.4,  1.6])
>>> np.around([.5, 1.5, 2.5, 3.5, 4.5]) # rounds to nearest even value
array([ 0.,  2.,  2.,  4.,  4.])
>>> np.around([1,2,3,11], decimals=1) # ndarray of ints is returned
array([ 1,  2,  3, 11])
>>> np.around([1,2,3,11], decimals=-1)
array([ 0,  0,  0, 10])
dask.array.array(object, dtype=None, copy=True, order='K', subok=False, ndmin=0)

Create an array.

Parameters:

object : array_like

An array, any object exposing the array interface, an object whose __array__ method returns an array, or any (nested) sequence.

dtype : data-type, optional

The desired data-type for the array. If not given, then the type will be determined as the minimum type required to hold the objects in the sequence. This argument can only be used to ‘upcast’ the array. For downcasting, use the .astype(t) method.

copy : bool, optional

If true (default), then the object is copied. Otherwise, a copy will only be made if __array__ returns a copy, if obj is a nested sequence, or if a copy is needed to satisfy any of the other requirements (dtype, order, etc.).

order : {‘K’, ‘A’, ‘C’, ‘F’}, optional

Specify the memory layout of the array. If object is not an array, the newly created array will be in C order (row major) unless ‘F’ is specified, in which case it will be in Fortran order (column major). If object is an array the following holds.

order no copy copy=True
‘K’ unchanged F & C order preserved, otherwise most similar order
‘A’ unchanged F order if input is F and not C, otherwise C order
‘C’ C order C order
‘F’ F order F order

When copy=False and a copy is made for other reasons, the result is the same as if copy=True, with some exceptions for A, see the Notes section. The default order is ‘K’.

subok : bool, optional

If True, then sub-classes will be passed-through, otherwise the returned array will be forced to be a base-class array (default).

ndmin : int, optional

Specifies the minimum number of dimensions that the resulting array should have. Ones will be pre-pended to the shape as needed to meet this requirement.
Returns:

out : ndarray

An array object satisfying the specified requirements.

See also

empty, empty_like, zeros, zeros_like, ones, ones_like, full, full_like

Notes

When order is ‘A’ and object is an array in neither ‘C’ nor ‘F’ order, and a copy is forced by a change in dtype, then the order of the result is not necessarily ‘C’ as expected. This is likely a bug.

Examples

>>> np.array([1, 2, 3])
array([1, 2, 3])

Upcasting:

>>> np.array([1, 2, 3.0])
array([ 1.,  2.,  3.])

More than one dimension:

>>> np.array([[1, 2], [3, 4]])
array([[1, 2],
       [3, 4]])

Minimum dimensions 2:

>>> np.array([1, 2, 3], ndmin=2)
array([[1, 2, 3]])

Type provided:

>>> np.array([1, 2, 3], dtype=complex)
array([ 1.+0.j,  2.+0.j,  3.+0.j])

Data-type consisting of more than one element:

>>> x = np.array([(1,2),(3,4)],dtype=[('a','<i4'),('b','<i4')])
>>> x['a']
array([1, 3])

Creating an array from sub-classes:

>>> np.array(np.mat('1 2; 3 4'))
array([[1, 2],
       [3, 4]])

>>> np.array(np.mat('1 2; 3 4'), subok=True)
matrix([[1, 2],
        [3, 4]])
dask.array.bincount(x, weights=None, minlength=None)

Count number of occurrences of each value in array of non-negative ints.

The number of bins (of size 1) is one larger than the largest value in x. If minlength is specified, there will be at least this number of bins in the output array (though it will be longer if necessary, depending on the contents of x). Each bin gives the number of occurrences of its index value in x. If weights is specified the input array is weighted by it, i.e. if a value n is found at position i, out[n] += weight[i] instead of out[n] += 1.

Parameters:

x : array_like, 1 dimension, nonnegative ints

Input array.

weights : array_like, optional

Weights, array of the same shape as x.

minlength : int, optional

A minimum number of bins for the output array.

New in version 1.6.0.
Returns:

out : ndarray of ints

The result of binning the input array. The length of out is equal to np.amax(x)+1.
Raises:

ValueError

If the input is not 1-dimensional, or contains elements with negative values, or if minlength is non-positive.

TypeError

If the type of the input is float or complex.

See also

histogram, digitize, unique

Examples

>>> np.bincount(np.arange(5))
array([1, 1, 1, 1, 1])
>>> np.bincount(np.array([0, 1, 1, 3, 2, 1, 7]))
array([1, 3, 1, 1, 0, 0, 0, 1])

>>> x = np.array([0, 1, 1, 3, 2, 1, 7, 23])
>>> np.bincount(x).size == np.amax(x)+1
True

The input array needs to be of integer dtype, otherwise a TypeError is raised:

>>> np.bincount(np.arange(5, dtype=np.float))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: array cannot be safely cast to required type

A possible use of bincount is to perform sums over variable-size chunks of an array, using the weights keyword.

>>> w = np.array([0.3, 0.5, 0.2, 0.7, 1., -0.6]) # weights
>>> x = np.array([0, 1, 1, 2, 2, 2])
>>> np.bincount(x,  weights=w)
array([ 0.3,  0.7,  1.1])
dask.array.broadcast_to(array, shape, subok=False)

Broadcast an array to a new shape.

Parameters:

array : array_like

The array to broadcast.

shape : tuple

The shape of the desired array.

subok : bool, optional

If True, then sub-classes will be passed-through, otherwise the returned array will be forced to be a base-class array (default).
Returns:

broadcast : array

A readonly view on the original array with the given shape. It is typically not contiguous. Furthermore, more than one element of a broadcasted array may refer to a single memory location.
Raises:

ValueError

If the array is not compatible with the new shape according to NumPy’s broadcasting rules.

Notes

New in version 1.10.0.

Examples

>>> x = np.array([1, 2, 3])
>>> np.broadcast_to(x, (3, 3))
array([[1, 2, 3],
       [1, 2, 3],
       [1, 2, 3]])
dask.array.coarsen(reduction, x, axes, trim_excess=False)

Coarsen array by applying reduction to fixed size neighborhoods

Parameters:

reduction: function

Function like np.sum, np.mean, etc...

x: np.ndarray

Array to be coarsened

axes: dict

Mapping of axis to coarsening factor

Examples

>>> x = np.array([1, 2, 3, 4, 5, 6])
>>> coarsen(np.sum, x, {0: 2})
array([ 3,  7, 11])
>>> coarsen(np.max, x, {0: 3})
array([3, 6])

Provide dictionary of scale per dimension

>>> x = np.arange(24).reshape((4, 6))
>>> x
array([[ 0,  1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17],
       [18, 19, 20, 21, 22, 23]])

>>> coarsen(np.min, x, {0: 2, 1: 3})
array([[ 0,  3],
       [12, 15]])

You must avoid excess elements explicitly

>>> x = np.array([1, 2, 3, 4, 5, 6, 7, 8])
>>> coarsen(np.min, x, {0: 3}, trim_excess=True)
array([1, 4])
dask.array.ceil(x[, out])

Return the ceiling of the input, element-wise.

The ceil of the scalar x is the smallest integer i, such that i >= x. It is often denoted as \(\lceil x \rceil\).

Parameters:

x : array_like

Input data.
Returns:

y : ndarray or scalar

The ceiling of each element in x, with float dtype.

See also

floor, trunc, rint

Examples

>>> a = np.array([-1.7, -1.5, -0.2, 0.2, 1.5, 1.7, 2.0])  
>>> np.ceil(a)  
array([-1., -1., -0.,  1.,  2.,  2.,  2.])
dask.array.choose(a, choices, out=None, mode='raise')

Construct an array from an index array and a set of arrays to choose from.

First of all, if confused or uncertain, definitely look at the Examples - in its full generality, this function is less simple than it might seem from the following code description (below ndi = numpy.lib.index_tricks):

np.choose(a,c) == np.array([c[a[I]][I] for I in ndi.ndindex(a.shape)]).

But this omits some subtleties. Here is a fully general summary:

Given an “index” array (a) of integers and a sequence of n arrays (choices), a and each choice array are first broadcast, as necessary, to arrays of a common shape; calling these Ba and Bchoices[i], i = 0,...,n-1 we have that, necessarily, Ba.shape == Bchoices[i].shape for each i. Then, a new array with shape Ba.shape is created as follows:

  • if mode=raise (the default), then, first of all, each element of a (and thus Ba) must be in the range [0, n-1]; now, suppose that i (in that range) is the value at the (j0, j1, ..., jm) position in Ba - then the value at the same position in the new array is the value in Bchoices[i] at that same position;
  • if mode=wrap, values in a (and thus Ba) may be any (signed) integer; modular arithmetic is used to map integers outside the range [0, n-1] back into that range; and then the new array is constructed as above;
  • if mode=clip, values in a (and thus Ba) may be any (signed) integer; negative integers are mapped to 0; values greater than n-1 are mapped to n-1; and then the new array is constructed as above.
Parameters:

a : int array

This array must contain integers in [0, n-1], where n is the number of choices, unless mode=wrap or mode=clip, in which cases any integers are permissible.

choices : sequence of arrays

Choice arrays. a and all of the choices must be broadcastable to the same shape. If choices is itself an array (not recommended), then its outermost dimension (i.e., the one corresponding to choices.shape[0]) is taken as defining the “sequence”.

out : array, optional

If provided, the result will be inserted into this array. It should be of the appropriate shape and dtype.

mode : {‘raise’ (default), ‘wrap’, ‘clip’}, optional

Specifies how indices outside [0, n-1] will be treated:

  • ‘raise’ : an exception is raised
  • ‘wrap’ : value becomes value mod n
  • ‘clip’ : values < 0 are mapped to 0, values > n-1 are mapped to n-1
Returns:

merged_array : array

The merged result.
Raises:

ValueError: shape mismatch

If a and each choice array are not all broadcastable to the same shape.

See also

ndarray.choose
equivalent method

Notes

To reduce the chance of misinterpretation, even though the following “abuse” is nominally supported, choices should neither be, nor be thought of as, a single array, i.e., the outermost sequence-like container should be either a list or a tuple.

Examples

>>> choices = [[0, 1, 2, 3], [10, 11, 12, 13],
...   [20, 21, 22, 23], [30, 31, 32, 33]]
>>> np.choose([2, 3, 1, 0], choices
... # the first element of the result will be the first element of the
... # third (2+1) "array" in choices, namely, 20; the second element
... # will be the second element of the fourth (3+1) choice array, i.e.,
... # 31, etc.
... )
array([20, 31, 12,  3])
>>> np.choose([2, 4, 1, 0], choices, mode='clip') # 4 goes to 3 (4-1)
array([20, 31, 12,  3])
>>> # because there are 4 choice arrays
>>> np.choose([2, 4, 1, 0], choices, mode='wrap') # 4 goes to (4 mod 4)
array([20,  1, 12,  3])
>>> # i.e., 0

A couple examples illustrating how choose broadcasts:

>>> a = [[1, 0, 1], [0, 1, 0], [1, 0, 1]]
>>> choices = [-10, 10]
>>> np.choose(a, choices)
array([[ 10, -10,  10],
       [-10,  10, -10],
       [ 10, -10,  10]])

>>> # With thanks to Anne Archibald
>>> a = np.array([0, 1]).reshape((2,1,1))
>>> c1 = np.array([1, 2, 3]).reshape((1,3,1))
>>> c2 = np.array([-1, -2, -3, -4, -5]).reshape((1,1,5))
>>> np.choose(a, (c1, c2)) # result is 2x3x5, res[0,:,:]=c1, res[1,:,:]=c2
array([[[ 1,  1,  1,  1,  1],
        [ 2,  2,  2,  2,  2],
        [ 3,  3,  3,  3,  3]],
       [[-1, -2, -3, -4, -5],
        [-1, -2, -3, -4, -5],
        [-1, -2, -3, -4, -5]]])
dask.array.clip(*args, **kwargs)

Clip (limit) the values in an array.

Given an interval, values outside the interval are clipped to the interval edges. For example, if an interval of [0, 1] is specified, values smaller than 0 become 0, and values larger than 1 become 1.

Parameters:

a : array_like

Array containing elements to clip.

a_min : scalar or array_like

Minimum value.

a_max : scalar or array_like

Maximum value. If a_min or a_max are array_like, then they will be broadcasted to the shape of a.

out : ndarray, optional

The results will be placed in this array. It may be the input array for in-place clipping. out must be of the right shape to hold the output. Its type is preserved.
Returns:

clipped_array : ndarray

An array with the elements of a, but where values < a_min are replaced with a_min, and those > a_max with a_max.

See also

numpy.doc.ufuncs
Section “Output arguments”

Examples

>>> a = np.arange(10)  
>>> np.clip(a, 1, 8)  
array([1, 1, 2, 3, 4, 5, 6, 7, 8, 8])
>>> a  
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
>>> np.clip(a, 3, 6, out=a)  
array([3, 3, 3, 3, 4, 5, 6, 6, 6, 6])
>>> a = np.arange(10)  
>>> a  
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
>>> np.clip(a, [3,4,1,1,1,4,4,4,4,4], 8)  
array([3, 4, 2, 3, 4, 5, 6, 7, 8, 8])
dask.array.compress(condition, a, axis=None, out=None)

Return selected slices of an array along given axis.

When working along a given axis, a slice along that axis is returned in output for each index where condition evaluates to True. When working on a 1-D array, compress is equivalent to extract.

Parameters:

condition : 1-D array of bools

Array that selects which entries to return. If len(condition) is less than the size of a along the given axis, then output is truncated to the length of the condition array.

a : array_like

Array from which to extract a part.

axis : int, optional

Axis along which to take slices. If None (default), work on the flattened array.

out : ndarray, optional

Output array. Its type is preserved and it must be of the right shape to hold the output.
Returns:

compressed_array : ndarray

A copy of a without the slices along axis for which condition is false.

See also

take, choose, diag, diagonal, select

ndarray.compress
Equivalent method in ndarray
np.extract
Equivalent method when working on 1-D arrays
numpy.doc.ufuncs
Section “Output arguments”

Examples

>>> a = np.array([[1, 2], [3, 4], [5, 6]])
>>> a
array([[1, 2],
       [3, 4],
       [5, 6]])
>>> np.compress([0, 1], a, axis=0)
array([[3, 4]])
>>> np.compress([False, True, True], a, axis=0)
array([[3, 4],
       [5, 6]])
>>> np.compress([False, True], a, axis=1)
array([[2],
       [4],
       [6]])

Working on the flattened array does not return slices along an axis but selects elements.

>>> np.compress([False, True], a)
array([2])
dask.array.concatenate(seq, axis=0, allow_unknown_chunksizes=False)

Concatenate arrays along an existing axis

Given a sequence of dask Arrays form a new dask Array by stacking them along an existing dimension (axis=0 by default)

Parameters:

seq: list of dask.arrays

axis: int

Dimension along which to align all of the arrays

allow_unknown_chunksizes: bool

Allow unknown chunksizes, such as come from converting from dask dataframes. Dask.array is unable to verify that chunks line up. If data comes from differently aligned sources then this can cause unexpected results.

See also

stack

Examples

Create slices

>>> import dask.array as da
>>> import numpy as np

>>> data = [from_array(np.ones((4, 4)), chunks=(2, 2))
...          for i in range(3)]

>>> x = da.concatenate(data, axis=0)
>>> x.shape
(12, 4)

>>> da.concatenate(data, axis=1).shape
(4, 12)

Result is a new dask Array

dask.array.conj(x[, out])

Return the complex conjugate, element-wise.

The complex conjugate of a complex number is obtained by changing the sign of its imaginary part.

Parameters:

x : array_like

Input value.
Returns:

y : ndarray

The complex conjugate of x, with same dtype as y.

Examples

>>> np.conjugate(1+2j)  
(1-2j)

>>> x = np.eye(2) + 1j * np.eye(2)  
>>> np.conjugate(x)  
array([[ 1.-1.j,  0.-0.j],
       [ 0.-0.j,  1.-1.j]])
dask.array.copysign(x1, x2[, out])

Change the sign of x1 to that of x2, element-wise.

If both arguments are arrays or sequences, they have to be of the same length. If x2 is a scalar, its sign will be copied to all elements of x1.

Parameters:

x1 : array_like

Values to change the sign of.

x2 : array_like

The sign of x2 is copied to x1.

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

out : array_like

The values of x1 with the sign of x2.

Examples

>>> np.copysign(1.3, -1)  
-1.3
>>> 1/np.copysign(0, 1)  
inf
>>> 1/np.copysign(0, -1)  
-inf

>>> np.copysign([-1, 0, 1], -1.1)  
array([-1., -0., -1.])
>>> np.copysign([-1, 0, 1], np.arange(3)-1)  
array([-1.,  0.,  1.])
dask.array.corrcoef(x, y=None, rowvar=1, bias=<class 'numpy._globals._NoValue'>, ddof=<class 'numpy._globals._NoValue'>)

Return Pearson product-moment correlation coefficients.

Please refer to the documentation for cov for more detail. The relationship between the correlation coefficient matrix, R, and the covariance matrix, C, is

\[R_{ij} = \frac{ C_{ij} } { \sqrt{ C_{ii} * C_{jj} } }\]

The values of R are between -1 and 1, inclusive.

Parameters:

x : array_like

A 1-D or 2-D array containing multiple variables and observations. Each row of x represents a variable, and each column a single observation of all those variables. Also see rowvar below.

y : array_like, optional

An additional set of variables and observations. y has the same shape as x.

rowvar : int, optional

If rowvar is non-zero (default), then each row represents a variable, with observations in the columns. Otherwise, the relationship is transposed: each column represents a variable, while the rows contain observations.

bias : _NoValue, optional

Has no effect, do not use.

Deprecated since version 1.10.0.

ddof : _NoValue, optional

Has no effect, do not use.

Deprecated since version 1.10.0.
Returns:

R : ndarray

The correlation coefficient matrix of the variables.

See also

cov
Covariance matrix

Notes

Due to floating point rounding the resulting array may not be Hermitian, the diagonal elements may not be 1, and the elements may not satisfy the inequality abs(a) <= 1. The real and imaginary parts are clipped to the interval [-1, 1] in an attempt to improve on that situation but is not much help in the complex case.

This function accepts but discards arguments bias and ddof. This is for backwards compatibility with previous versions of this function. These arguments had no effect on the return values of the function and can be safely ignored in this and previous versions of numpy.

dask.array.cos(x[, out])

Cosine element-wise.

Parameters:

x : array_like

Input array in radians.

out : ndarray, optional

Output array of same shape as x.
Returns:

y : ndarray

The corresponding cosine values.
Raises:

ValueError: invalid return array shape

if out is provided and out.shape != x.shape (See Examples)

Notes

If out is provided, the function writes the result into it, and returns a reference to out. (See Examples)

References

M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions. New York, NY: Dover, 1972.

Examples

>>> np.cos(np.array([0, np.pi/2, np.pi]))  
array([  1.00000000e+00,   6.12303177e-17,  -1.00000000e+00])
>>>
>>> # Example of providing the optional output parameter
>>> out2 = np.cos([0.1], out1)  
>>> out2 is out1  
True
>>>
>>> # Example of ValueError due to provision of shape mis-matched `out`
>>> np.cos(np.zeros((3,3)),np.zeros((2,2)))  
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: invalid return array shape
dask.array.cosh(x[, out])

Hyperbolic cosine, element-wise.

Equivalent to 1/2 * (np.exp(x) + np.exp(-x)) and np.cos(1j*x).

Parameters:

x : array_like

Input array.
Returns:

out : ndarray

Output array of same shape as x.

Examples

>>> np.cosh(0)  
1.0

The hyperbolic cosine describes the shape of a hanging cable:

>>> import matplotlib.pyplot as plt  
>>> x = np.linspace(-4, 4, 1000)  
>>> plt.plot(x, np.cosh(x))  
>>> plt.show()
dask.array.cov(m, y=None, rowvar=True, bias=False, ddof=None, fweights=None, aweights=None)

Estimate a covariance matrix, given data and weights.

Covariance indicates the level to which two variables vary together. If we examine N-dimensional samples, \(X = [x_1, x_2, ... x_N]^T\), then the covariance matrix element \(C_{ij}\) is the covariance of \(x_i\) and \(x_j\). The element \(C_{ii}\) is the variance of \(x_i\).

See the notes for an outline of the algorithm.

Parameters:

m : array_like

A 1-D or 2-D array containing multiple variables and observations. Each row of m represents a variable, and each column a single observation of all those variables. Also see rowvar below.

y : array_like, optional

An additional set of variables and observations. y has the same form as that of m.

rowvar : bool, optional

If rowvar is True (default), then each row represents a variable, with observations in the columns. Otherwise, the relationship is transposed: each column represents a variable, while the rows contain observations.

bias : bool, optional

Default normalization (False) is by (N - 1), where N is the number of observations given (unbiased estimate). If bias is True, then normalization is by N. These values can be overridden by using the keyword ddof in numpy versions >= 1.5.

ddof : int, optional

If not None the default value implied by bias is overridden. Note that ddof=1 will return the unbiased estimate, even if both fweights and aweights are specified, and ddof=0 will return the simple average. See the notes for the details. The default value is None.

New in version 1.5.

fweights : array_like, int, optional

1-D array of integer freguency weights; the number of times each observation vector should be repeated.

New in version 1.10.

aweights : array_like, optional

1-D array of observation vector weights. These relative weights are typically large for observations considered “important” and smaller for observations considered less “important”. If ddof=0 the array of weights can be used to assign probabilities to observation vectors.

New in version 1.10.
Returns:

out : ndarray

The covariance matrix of the variables.

See also

corrcoef
Normalized covariance matrix

Notes

Assume that the observations are in the columns of the observation array m and let f = fweights and a = aweights for brevity. The steps to compute the weighted covariance are as follows:

>>> w = f * a
>>> v1 = np.sum(w)
>>> v2 = np.sum(w * a)
>>> m -= np.sum(m * w, axis=1, keepdims=True) / v1
>>> cov = np.dot(m * w, m.T) * v1 / (v1**2 - ddof * v2)

Note that when a == 1, the normalization factor v1 / (v1**2 - ddof * v2) goes over to 1 / (np.sum(f) - ddof) as it should.

Examples

Consider two variables, \(x_0\) and \(x_1\), which correlate perfectly, but in opposite directions:

>>> x = np.array([[0, 2], [1, 1], [2, 0]]).T
>>> x
array([[0, 1, 2],
       [2, 1, 0]])

Note how \(x_0\) increases while \(x_1\) decreases. The covariance matrix shows this clearly:

>>> np.cov(x)
array([[ 1., -1.],
       [-1.,  1.]])

Note that element \(C_{0,1}\), which shows the correlation between \(x_0\) and \(x_1\), is negative.

Further, note how x and y are combined:

>>> x = [-2.1, -1,  4.3]
>>> y = [3,  1.1,  0.12]
>>> X = np.vstack((x,y))
>>> print(np.cov(X))
[[ 11.71        -4.286     ]
 [ -4.286        2.14413333]]
>>> print(np.cov(x, y))
[[ 11.71        -4.286     ]
 [ -4.286        2.14413333]]
>>> print(np.cov(x))
11.71
dask.array.cumprod(a, axis=None, dtype=None, out=None)

Return the cumulative product of elements along a given axis.

Parameters:

a : array_like

Input array.

axis : int, optional

Axis along which the cumulative product is computed. By default the input is flattened.

dtype : dtype, optional

Type of the returned array, as well as of the accumulator in which the elements are multiplied. If dtype is not specified, it defaults to the dtype of a, unless a has an integer dtype with a precision less than that of the default platform integer. In that case, the default platform integer is used instead.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape and buffer length as the expected output but the type of the resulting values will be cast if necessary.
Returns:

cumprod : ndarray

A new array holding the result is returned unless out is specified, in which case a reference to out is returned.

See also

numpy.doc.ufuncs
Section “Output arguments”

Notes

Arithmetic is modular when using integer types, and no error is raised on overflow.

Examples

>>> a = np.array([1,2,3])
>>> np.cumprod(a) # intermediate results 1, 1*2
...               # total product 1*2*3 = 6
array([1, 2, 6])
>>> a = np.array([[1, 2, 3], [4, 5, 6]])
>>> np.cumprod(a, dtype=float) # specify type of output
array([   1.,    2.,    6.,   24.,  120.,  720.])

The cumulative product for each column (i.e., over the rows) of a:

>>> np.cumprod(a, axis=0)
array([[ 1,  2,  3],
       [ 4, 10, 18]])

The cumulative product for each row (i.e. over the columns) of a:

>>> np.cumprod(a,axis=1)
array([[  1,   2,   6],
       [  4,  20, 120]])
dask.array.cumsum(a, axis=None, dtype=None, out=None)

Return the cumulative sum of the elements along a given axis.

Parameters:

a : array_like

Input array.

axis : int, optional

Axis along which the cumulative sum is computed. The default (None) is to compute the cumsum over the flattened array.

dtype : dtype, optional

Type of the returned array and of the accumulator in which the elements are summed. If dtype is not specified, it defaults to the dtype of a, unless a has an integer dtype with a precision less than that of the default platform integer. In that case, the default platform integer is used.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape and buffer length as the expected output but the type will be cast if necessary. See doc.ufuncs (Section “Output arguments”) for more details.
Returns:

cumsum_along_axis : ndarray.

A new array holding the result is returned unless out is specified, in which case a reference to out is returned. The result has the same size as a, and the same shape as a if axis is not None or a is a 1-d array.

See also

sum
Sum array elements.
trapz
Integration of array values using the composite trapezoidal rule.
diff
Calculate the n-th discrete difference along given axis.

Notes

Arithmetic is modular when using integer types, and no error is raised on overflow.

Examples

>>> a = np.array([[1,2,3], [4,5,6]])
>>> a
array([[1, 2, 3],
       [4, 5, 6]])
>>> np.cumsum(a)
array([ 1,  3,  6, 10, 15, 21])
>>> np.cumsum(a, dtype=float)     # specifies type of output value(s)
array([  1.,   3.,   6.,  10.,  15.,  21.])

>>> np.cumsum(a,axis=0)      # sum over rows for each of the 3 columns
array([[1, 2, 3],
       [5, 7, 9]])
>>> np.cumsum(a,axis=1)      # sum over columns for each of the 2 rows
array([[ 1,  3,  6],
       [ 4,  9, 15]])
dask.array.deg2rad(x[, out])

Convert angles from degrees to radians.

Parameters:

x : array_like

Angles in degrees.
Returns:

y : ndarray

The corresponding angle in radians.

See also

rad2deg
Convert angles from radians to degrees.
unwrap
Remove large jumps in angle by wrapping.

Notes

New in version 1.3.0.

deg2rad(x) is x * pi / 180.

Examples

>>> np.deg2rad(180)  
3.1415926535897931
dask.array.degrees(x[, out])

Convert angles from radians to degrees.

Parameters:

x : array_like

Input array in radians.

out : ndarray, optional

Output array of same shape as x.
Returns:

y : ndarray of floats

The corresponding degree values; if out was supplied this is a reference to it.

See also

rad2deg
equivalent function

Examples

Convert a radian array to degrees

>>> rad = np.arange(12.)*np.pi/6  
>>> np.degrees(rad)  
array([   0.,   30.,   60.,   90.,  120.,  150.,  180.,  210.,  240.,
        270.,  300.,  330.])

>>> out = np.zeros((rad.shape))  
>>> r = degrees(rad, out)  
>>> np.all(r == out)  
True
dask.array.diag(v, k=0)

Extract a diagonal or construct a diagonal array.

See the more detailed documentation for numpy.diagonal if you use this function to extract a diagonal and wish to write to the resulting array; whether it returns a copy or a view depends on what version of numpy you are using.

Parameters:

v : array_like

If v is a 2-D array, return a copy of its k-th diagonal. If v is a 1-D array, return a 2-D array with v on the k-th diagonal.

k : int, optional

Diagonal in question. The default is 0. Use k>0 for diagonals above the main diagonal, and k<0 for diagonals below the main diagonal.
Returns:

out : ndarray

The extracted diagonal or constructed diagonal array.

See also

diagonal
Return specified diagonals.
diagflat
Create a 2-D array with the flattened input as a diagonal.
trace
Sum along diagonals.
triu
Upper triangle of an array.
tril
Lower triangle of an array.

Examples

>>> x = np.arange(9).reshape((3,3))
>>> x
array([[0, 1, 2],
       [3, 4, 5],
       [6, 7, 8]])

>>> np.diag(x)
array([0, 4, 8])
>>> np.diag(x, k=1)
array([1, 5])
>>> np.diag(x, k=-1)
array([3, 7])

>>> np.diag(np.diag(x))
array([[0, 0, 0],
       [0, 4, 0],
       [0, 0, 8]])
dask.array.digitize(x, bins, right=False)

Return the indices of the bins to which each value in input array belongs.

Each index i returned is such that bins[i-1] <= x < bins[i] if bins is monotonically increasing, or bins[i-1] > x >= bins[i] if bins is monotonically decreasing. If values in x are beyond the bounds of bins, 0 or len(bins) is returned as appropriate. If right is True, then the right bin is closed so that the index i is such that bins[i-1] < x <= bins[i] or bins[i-1] >= x > bins[i]`` if bins is monotonically increasing or decreasing, respectively.

Parameters:

x : array_like

Input array to be binned. Prior to NumPy 1.10.0, this array had to be 1-dimensional, but can now have any shape.

bins : array_like

Array of bins. It has to be 1-dimensional and monotonic.

right : bool, optional

Indicating whether the intervals include the right or the left bin edge. Default behavior is (right==False) indicating that the interval does not include the right edge. The left bin end is open in this case, i.e., bins[i-1] <= x < bins[i] is the default behavior for monotonically increasing bins.
Returns:

out : ndarray of ints

Output array of indices, of same shape as x.
Raises:

ValueError

If bins is not monotonic.

TypeError

If the type of the input is complex.

See also

bincount, histogram, unique

Notes

If values in x are such that they fall outside the bin range, attempting to index bins with the indices that digitize returns will result in an IndexError.

New in version 1.10.0.

np.digitize is implemented in terms of np.searchsorted. This means that a binary search is used to bin the values, which scales much better for larger number of bins than the previous linear search. It also removes the requirement for the input array to be 1-dimensional.

Examples

>>> x = np.array([0.2, 6.4, 3.0, 1.6])
>>> bins = np.array([0.0, 1.0, 2.5, 4.0, 10.0])
>>> inds = np.digitize(x, bins)
>>> inds
array([1, 4, 3, 2])
>>> for n in range(x.size):
...   print(bins[inds[n]-1], "<=", x[n], "<", bins[inds[n]])
...
0.0 <= 0.2 < 1.0
4.0 <= 6.4 < 10.0
2.5 <= 3.0 < 4.0
1.0 <= 1.6 < 2.5

>>> x = np.array([1.2, 10.0, 12.4, 15.5, 20.])
>>> bins = np.array([0, 5, 10, 15, 20])
>>> np.digitize(x,bins,right=True)
array([1, 2, 3, 4, 4])
>>> np.digitize(x,bins,right=False)
array([1, 3, 3, 4, 5])
dask.array.dot(a, b, out=None)

Dot product of two arrays.

For 2-D arrays it is equivalent to matrix multiplication, and for 1-D arrays to inner product of vectors (without complex conjugation). For N dimensions it is a sum product over the last axis of a and the second-to-last of b:

dot(a, b)[i,j,k,m] = sum(a[i,j,:] * b[k,:,m])
Parameters:

a : array_like

First argument.

b : array_like

Second argument.

out : ndarray, optional

Output argument. This must have the exact kind that would be returned if it was not used. In particular, it must have the right type, must be C-contiguous, and its dtype must be the dtype that would be returned for dot(a,b). This is a performance feature. Therefore, if these conditions are not met, an exception is raised, instead of attempting to be flexible.
Returns:

output : ndarray

Returns the dot product of a and b. If a and b are both scalars or both 1-D arrays then a scalar is returned; otherwise an array is returned. If out is given, then it is returned.
Raises:

ValueError

If the last dimension of a is not the same size as the second-to-last dimension of b.

See also

vdot
Complex-conjugating dot product.
tensordot
Sum products over arbitrary axes.
einsum
Einstein summation convention.
matmul
‘@’ operator as method with out parameter.

Examples

>>> np.dot(3, 4)
12

Neither argument is complex-conjugated:

>>> np.dot([2j, 3j], [2j, 3j])
(-13+0j)

For 2-D arrays it is the matrix product:

>>> a = [[1, 0], [0, 1]]
>>> b = [[4, 1], [2, 2]]
>>> np.dot(a, b)
array([[4, 1],
       [2, 2]])

>>> a = np.arange(3*4*5*6).reshape((3,4,5,6))
>>> b = np.arange(3*4*5*6)[::-1].reshape((5,4,6,3))
>>> np.dot(a, b)[2,3,2,1,2,2]
499128
>>> sum(a[2,3,2,:] * b[1,2,:,2])
499128
dask.array.dstack(tup)

Stack arrays in sequence depth wise (along third axis).

Takes a sequence of arrays and stack them along the third axis to make a single array. Rebuilds arrays divided by dsplit. This is a simple way to stack 2D arrays (images) into a single 3D array for processing.

This function continues to be supported for backward compatibility, but you should prefer np.concatenate or np.stack. The np.stack function was added in NumPy 1.10.

Parameters:

tup : sequence of arrays

Arrays to stack. All of them must have the same shape along all but the third axis.
Returns:

stacked : ndarray

The array formed by stacking the given arrays.

See also

stack
Join a sequence of arrays along a new axis.
vstack
Stack along first axis.
hstack
Stack along second axis.
concatenate
Join a sequence of arrays along an existing axis.
dsplit
Split array along third axis.

Notes

Equivalent to np.concatenate(tup, axis=2).

Examples

>>> a = np.array((1,2,3))
>>> b = np.array((2,3,4))
>>> np.dstack((a,b))
array([[[1, 2],
        [2, 3],
        [3, 4]]])

>>> a = np.array([[1],[2],[3]])
>>> b = np.array([[2],[3],[4]])
>>> np.dstack((a,b))
array([[[1, 2]],
       [[2, 3]],
       [[3, 4]]])
dask.array.empty()

Blocked variant of empty

Follows the signature of empty exactly except that it also requires a keyword argument chunks=(...)

Original signature follows below. empty(shape, dtype=float, order=’C’)

Return a new array of given shape and type, without initializing entries.

Parameters:

shape : int or tuple of int

Shape of the empty array

dtype : data-type, optional

Desired output data-type.

order : {‘C’, ‘F’}, optional

Whether to store multi-dimensional data in row-major (C-style) or column-major (Fortran-style) order in memory.
Returns:

out : ndarray

Array of uninitialized (arbitrary) data of the given shape, dtype, and order. Object arrays will be initialized to None.

See also

empty_like, zeros, ones

Notes

empty, unlike zeros, does not set the array values to zero, and may therefore be marginally faster. On the other hand, it requires the user to manually set all the values in the array, and should be used with caution.

Examples

>>> np.empty([2, 2])
array([[ -9.74499359e+001,   6.69583040e-309],
       [  2.13182611e-314,   3.06959433e-309]])         #random

>>> np.empty([2, 2], dtype=int)
array([[-1073741821, -1067949133],
       [  496041986,    19249760]])                     #random
dask.array.exp(x[, out])

Calculate the exponential of all elements in the input array.

Parameters:

x : array_like

Input values.
Returns:

out : ndarray

Output array, element-wise exponential of x.

See also

expm1
Calculate exp(x) - 1 for all elements in the array.
exp2
Calculate 2**x for all elements in the array.

Notes

The irrational number e is also known as Euler’s number. It is approximately 2.718281, and is the base of the natural logarithm, ln (this means that, if \(x = \ln y = \log_e y\), then \(e^x = y\). For real input, exp(x) is always positive.

For complex arguments, x = a + ib, we can write \(e^x = e^a e^{ib}\). The first term, \(e^a\), is already known (it is the real argument, described above). The second term, \(e^{ib}\), is \(\cos b + i \sin b\), a function with magnitude 1 and a periodic phase.

References

[R225226]Wikipedia, “Exponential function”, http://en.wikipedia.org/wiki/Exponential_function
[R226226]M. Abramovitz and I. A. Stegun, “Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables,” Dover, 1964, p. 69, http://www.math.sfu.ca/~cbm/aands/page_69.htm

Examples

Plot the magnitude and phase of exp(x) in the complex plane:

>>> import matplotlib.pyplot as plt  

>>> x = np.linspace(-2*np.pi, 2*np.pi, 100)  
>>> xx = x + 1j * x[:, np.newaxis] # a + ib over complex plane  
>>> out = np.exp(xx)  

>>> plt.subplot(121)  
>>> plt.imshow(np.abs(out),  
...            extent=[-2*np.pi, 2*np.pi, -2*np.pi, 2*np.pi])
>>> plt.title('Magnitude of exp(x)')  

>>> plt.subplot(122)  
>>> plt.imshow(np.angle(out),  
...            extent=[-2*np.pi, 2*np.pi, -2*np.pi, 2*np.pi])
>>> plt.title('Phase (angle) of exp(x)')  
>>> plt.show()
dask.array.expm1(x[, out])

Calculate exp(x) - 1 for all elements in the array.

Parameters:

x : array_like

Input values.
Returns:

out : ndarray

Element-wise exponential minus one: out = exp(x) - 1.

See also

log1p
log(1 + x), the inverse of expm1.

Notes

This function provides greater precision than exp(x) - 1 for small values of x.

Examples

The true value of exp(1e-10) - 1 is 1.00000000005e-10 to about 32 significant digits. This example shows the superiority of expm1 in this case.

>>> np.expm1(1e-10)  
1.00000000005e-10
>>> np.exp(1e-10) - 1  
1.000000082740371e-10
dask.array.eye(N, chunks, M=None, k=0, dtype=<class 'float'>)

Return a 2-D Array with ones on the diagonal and zeros elsewhere.

Parameters:

N : int

Number of rows in the output.

chunks: int

chunk size of resulting blocks

M : int, optional

Number of columns in the output. If None, defaults to N.

k : int, optional

Index of the diagonal: 0 (the default) refers to the main diagonal, a positive value refers to an upper diagonal, and a negative value to a lower diagonal.

dtype : data-type, optional

Data-type of the returned array.
Returns:

I : Array of shape (N,M)

An array where all elements are equal to zero, except for the k-th diagonal, whose values are equal to one.
dask.array.fabs(x[, out])

Compute the absolute values element-wise.

This function returns the absolute values (positive magnitude) of the data in x. Complex values are not handled, use absolute to find the absolute values of complex data.

Parameters:

x : array_like

The array of numbers for which the absolute values are required. If x is a scalar, the result y will also be a scalar.

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

y : ndarray or scalar

The absolute values of x, the returned values are always floats.

See also

absolute
Absolute values including complex types.

Examples

>>> np.fabs(-1)  
1.0
>>> np.fabs([-1.2, 1.2])  
array([ 1.2,  1.2])
dask.array.fix(*args, **kwargs)

Round to nearest integer towards zero.

Round an array of floats element-wise to nearest integer towards zero. The rounded values are returned as floats.

Parameters:

x : array_like

An array of floats to be rounded

y : ndarray, optional

Output array
Returns:

out : ndarray of floats

The array of rounded numbers

See also

trunc, floor, ceil

around
Round to given number of decimals

Examples

>>> np.fix(3.14)  
3.0
>>> np.fix(3)  
3.0
>>> np.fix([2.1, 2.9, -2.1, -2.9])  
array([ 2.,  2., -2., -2.])
dask.array.floor(x[, out])

Return the floor of the input, element-wise.

The floor of the scalar x is the largest integer i, such that i <= x. It is often denoted as \(\lfloor x \rfloor\).

Parameters:

x : array_like

Input data.
Returns:

y : ndarray or scalar

The floor of each element in x.

See also

ceil, trunc, rint

Notes

Some spreadsheet programs calculate the “floor-towards-zero”, in other words floor(-2.5) == -2. NumPy instead uses the definition of floor where floor(-2.5) == -3.

Examples

>>> a = np.array([-1.7, -1.5, -0.2, 0.2, 1.5, 1.7, 2.0])  
>>> np.floor(a)  
array([-2., -2., -1.,  0.,  1.,  1.,  2.])
dask.array.fmax(x1, x2[, out])

Element-wise maximum of array elements.

Compare two arrays and returns a new array containing the element-wise maxima. If one of the elements being compared is a NaN, then the non-nan element is returned. If both elements are NaNs then the first is returned. The latter distinction is important for complex NaNs, which are defined as at least one of the real or imaginary parts being a NaN. The net effect is that NaNs are ignored when possible.

Parameters:

x1, x2 : array_like

The arrays holding the elements to be compared. They must have the same shape.
Returns:

y : ndarray or scalar

The maximum of x1 and x2, element-wise. Returns scalar if both x1 and x2 are scalars.

See also

fmin
Element-wise minimum of two arrays, ignores NaNs.
maximum
Element-wise maximum of two arrays, propagates NaNs.
amax
The maximum value of an array along a given axis, propagates NaNs.
nanmax
The maximum value of an array along a given axis, ignores NaNs.

minimum, amin, nanmin

Notes

New in version 1.3.0.

The fmax is equivalent to np.where(x1 >= x2, x1, x2) when neither x1 nor x2 are NaNs, but it is faster and does proper broadcasting.

Examples

>>> np.fmax([2, 3, 4], [1, 5, 2])  
array([ 2.,  5.,  4.])

>>> np.fmax(np.eye(2), [0.5, 2])  
array([[ 1. ,  2. ],
       [ 0.5,  2. ]])

>>> np.fmax([np.nan, 0, np.nan],[0, np.nan, np.nan])  
array([  0.,   0.,  NaN])
dask.array.fmin(x1, x2[, out])

Element-wise minimum of array elements.

Compare two arrays and returns a new array containing the element-wise minima. If one of the elements being compared is a NaN, then the non-nan element is returned. If both elements are NaNs then the first is returned. The latter distinction is important for complex NaNs, which are defined as at least one of the real or imaginary parts being a NaN. The net effect is that NaNs are ignored when possible.

Parameters:

x1, x2 : array_like

The arrays holding the elements to be compared. They must have the same shape.
Returns:

y : ndarray or scalar

The minimum of x1 and x2, element-wise. Returns scalar if both x1 and x2 are scalars.

See also

fmax
Element-wise maximum of two arrays, ignores NaNs.
minimum
Element-wise minimum of two arrays, propagates NaNs.
amin
The minimum value of an array along a given axis, propagates NaNs.
nanmin
The minimum value of an array along a given axis, ignores NaNs.

maximum, amax, nanmax

Notes

New in version 1.3.0.

The fmin is equivalent to np.where(x1 <= x2, x1, x2) when neither x1 nor x2 are NaNs, but it is faster and does proper broadcasting.

Examples

>>> np.fmin([2, 3, 4], [1, 5, 2])  
array([1, 3, 2])

>>> np.fmin(np.eye(2), [0.5, 2])  
array([[ 0.5,  0. ],
       [ 0. ,  1. ]])

>>> np.fmin([np.nan, 0, np.nan],[0, np.nan, np.nan])  
array([  0.,   0.,  NaN])
dask.array.fmod(x1, x2[, out])

Return the element-wise remainder of division.

This is the NumPy implementation of the C library function fmod, the remainder has the same sign as the dividend x1. It is equivalent to the Matlab(TM) rem function and should not be confused with the Python modulus operator x1 % x2.

Parameters:

x1 : array_like

Dividend.

x2 : array_like

Divisor.
Returns:

y : array_like

The remainder of the division of x1 by x2.

See also

remainder
Equivalent to the Python % operator.

divide

Notes

The result of the modulo operation for negative dividend and divisors is bound by conventions. For fmod, the sign of result is the sign of the dividend, while for remainder the sign of the result is the sign of the divisor. The fmod function is equivalent to the Matlab(TM) rem function.

Examples

>>> np.fmod([-3, -2, -1, 1, 2, 3], 2)  
array([-1,  0, -1,  1,  0,  1])
>>> np.remainder([-3, -2, -1, 1, 2, 3], 2)  
array([1, 0, 1, 1, 0, 1])

>>> np.fmod([5, 3], [2, 2.])  
array([ 1.,  1.])
>>> a = np.arange(-3, 3).reshape(3, 2)  
>>> a  
array([[-3, -2],
       [-1,  0],
       [ 1,  2]])
>>> np.fmod(a, [2,2])  
array([[-1,  0],
       [-1,  0],
       [ 1,  0]])
dask.array.frexp(x[, out1, out2])

Decompose the elements of x into mantissa and twos exponent.

Returns (mantissa, exponent), where x = mantissa * 2**exponent`. The mantissa is lies in the open interval(-1, 1), while the twos exponent is a signed integer.

Parameters:

x : array_like

Array of numbers to be decomposed.

out1 : ndarray, optional

Output array for the mantissa. Must have the same shape as x.

out2 : ndarray, optional

Output array for the exponent. Must have the same shape as x.
Returns:

(mantissa, exponent) : tuple of ndarrays, (float, int)

mantissa is a float array with values between -1 and 1. exponent is an int array which represents the exponent of 2.

See also

ldexp
Compute y = x1 * 2**x2, the inverse of frexp.

Notes

Complex dtypes are not supported, they will raise a TypeError.

Examples

>>> x = np.arange(9)  
>>> y1, y2 = np.frexp(x)  
>>> y1  
array([ 0.   ,  0.5  ,  0.5  ,  0.75 ,  0.5  ,  0.625,  0.75 ,  0.875,
        0.5  ])
>>> y2  
array([0, 1, 2, 2, 3, 3, 3, 3, 4])
>>> y1 * 2**y2  
array([ 0.,  1.,  2.,  3.,  4.,  5.,  6.,  7.,  8.])
dask.array.fromfunction(function, shape, **kwargs)

Construct an array by executing a function over each coordinate.

The resulting array therefore has a value fn(x, y, z) at coordinate (x, y, z).

Parameters:

function : callable

The function is called with N parameters, where N is the rank of shape. Each parameter represents the coordinates of the array varying along a specific axis. For example, if shape were (2, 2), then the parameters in turn be (0, 0), (0, 1), (1, 0), (1, 1).

shape : (N,) tuple of ints

Shape of the output array, which also determines the shape of the coordinate arrays passed to function.

dtype : data-type, optional

Data-type of the coordinate arrays passed to function. By default, dtype is float.
Returns:

fromfunction : any

The result of the call to function is passed back directly. Therefore the shape of fromfunction is completely determined by function. If function returns a scalar value, the shape of fromfunction would match the shape parameter.

See also

indices, meshgrid

Notes

Keywords other than dtype are passed to function.

Examples

>>> np.fromfunction(lambda i, j: i == j, (3, 3), dtype=int)
array([[ True, False, False],
       [False,  True, False],
       [False, False,  True]], dtype=bool)

>>> np.fromfunction(lambda i, j: i + j, (3, 3), dtype=int)
array([[0, 1, 2],
       [1, 2, 3],
       [2, 3, 4]])
dask.array.full(*args, **kwargs)

Blocked variant of full

Follows the signature of full exactly except that it also requires a keyword argument chunks=(...)

Original signature follows below.

Return a new array of given shape and type, filled with fill_value.

Parameters:

shape : int or sequence of ints

Shape of the new array, e.g., (2, 3) or 2.

fill_value : scalar

Fill value.

dtype : data-type, optional

The desired data-type for the array The default, None, means

np.array(fill_value).dtype.

order : {‘C’, ‘F’}, optional

Whether to store multidimensional data in C- or Fortran-contiguous (row- or column-wise) order in memory.
Returns:

out : ndarray

Array of fill_value with the given shape, dtype, and order.

See also

zeros_like
Return an array of zeros with shape and type of input.
ones_like
Return an array of ones with shape and type of input.
empty_like
Return an empty array with shape and type of input.
full_like
Fill an array with shape and type of input.
zeros
Return a new array setting values to zero.
ones
Return a new array setting values to one.
empty
Return a new uninitialized array.

Examples

>>> np.full((2, 2), np.inf)
array([[ inf,  inf],
       [ inf,  inf]])
>>> np.full((2, 2), 10)
array([[10, 10],
       [10, 10]])
dask.array.histogram(a, bins=None, range=None, normed=False, weights=None, density=None)

Blocked variant of numpy.histogram.

Follows the signature of numpy.histogram exactly with the following exceptions:

  • Either an iterable specifying the bins or the number of bins and a range argument is required as computing min and max over blocked arrays is an expensive operation that must be performed explicitly.
  • weights must be a dask.array.Array with the same block structure as a.

Examples

Using number of bins and range:

>>> import dask.array as da
>>> import numpy as np
>>> x = da.from_array(np.arange(10000), chunks=10)
>>> h, bins = da.histogram(x, bins=10, range=[0, 10000])
>>> bins
array([     0.,   1000.,   2000.,   3000.,   4000.,   5000.,   6000.,
         7000.,   8000.,   9000.,  10000.])
>>> h.compute()
array([1000, 1000, 1000, 1000, 1000, 1000, 1000, 1000, 1000, 1000])

Explicitly specifying the bins:

>>> h, bins = da.histogram(x, bins=np.array([0, 5000, 10000]))
>>> bins
array([    0,  5000, 10000])
>>> h.compute()
array([5000, 5000])
dask.array.hstack(tup)

Stack arrays in sequence horizontally (column wise).

Take a sequence of arrays and stack them horizontally to make a single array. Rebuild arrays divided by hsplit.

This function continues to be supported for backward compatibility, but you should prefer np.concatenate or np.stack. The np.stack function was added in NumPy 1.10.

Parameters:

tup : sequence of ndarrays

All arrays must have the same shape along all but the second axis.
Returns:

stacked : ndarray

The array formed by stacking the given arrays.

See also

stack
Join a sequence of arrays along a new axis.
vstack
Stack arrays in sequence vertically (row wise).
dstack
Stack arrays in sequence depth wise (along third axis).
concatenate
Join a sequence of arrays along an existing axis.
hsplit
Split array along second axis.

Notes

Equivalent to np.concatenate(tup, axis=1)

Examples

>>> a = np.array((1,2,3))
>>> b = np.array((2,3,4))
>>> np.hstack((a,b))
array([1, 2, 3, 2, 3, 4])
>>> a = np.array([[1],[2],[3]])
>>> b = np.array([[2],[3],[4]])
>>> np.hstack((a,b))
array([[1, 2],
       [2, 3],
       [3, 4]])
dask.array.hypot(x1, x2[, out])

Given the “legs” of a right triangle, return its hypotenuse.

Equivalent to sqrt(x1**2 + x2**2), element-wise. If x1 or x2 is scalar_like (i.e., unambiguously cast-able to a scalar type), it is broadcast for use with each element of the other argument. (See Examples)

Parameters:

x1, x2 : array_like

Leg of the triangle(s).

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

z : ndarray

The hypotenuse of the triangle(s).

Examples

>>> np.hypot(3*np.ones((3, 3)), 4*np.ones((3, 3)))  
array([[ 5.,  5.,  5.],
       [ 5.,  5.,  5.],
       [ 5.,  5.,  5.]])

Example showing broadcast of scalar_like argument:

>>> np.hypot(3*np.ones((3, 3)), [4])  
array([[ 5.,  5.,  5.],
       [ 5.,  5.,  5.],
       [ 5.,  5.,  5.]])
dask.array.imag(*args, **kwargs)

Return the imaginary part of the elements of the array.

Parameters:

val : array_like

Input array.
Returns:

out : ndarray

Output array. If val is real, the type of val is used for the output. If val has complex elements, the returned type is float.

See also

real, angle, real_if_close

Examples

>>> a = np.array([1+2j, 3+4j, 5+6j])  
>>> a.imag  
array([ 2.,  4.,  6.])
>>> a.imag = np.array([8, 10, 12])  
>>> a  
array([ 1. +8.j,  3.+10.j,  5.+12.j])
dask.array.indices(dimensions, dtype=<class 'int'>, chunks=None)

Implements NumPy’s indices for Dask Arrays.

Generates a grid of indices covering the dimensions provided.

The final array has the shape (len(dimensions), *dimensions). The chunks are used to specify the chunking for axis 1 up to len(dimensions). The 0th axis always has chunks of length 1.

Parameters:

dimensions : sequence of ints

The shape of the index grid.

dtype : dtype, optional

Type to use for the array. Default is int.

chunks : sequence of ints

The number of samples on each block. Note that the last block will have fewer samples if len(array) % chunks != 0.
Returns:

grid : dask array
dask.array.insert(arr, obj, values, axis=None)

Insert values along the given axis before the given indices.

Parameters:

arr : array_like

Input array.

obj : int, slice or sequence of ints

Object that defines the index or indices before which values is inserted.

New in version 1.8.0.

Support for multiple insertions when obj is a single scalar or a sequence with one element (similar to calling insert multiple times).

values : array_like

Values to insert into arr. If the type of values is different from that of arr, values is converted to the type of arr. values should be shaped so that arr[...,obj,...] = values is legal.

axis : int, optional

Axis along which to insert values. If axis is None then arr is flattened first.
Returns:

out : ndarray

A copy of arr with values inserted. Note that insert does not occur in-place: a new array is returned. If axis is None, out is a flattened array.

See also

append
Append elements at the end of an array.
concatenate
Join a sequence of arrays along an existing axis.
delete
Delete elements from an array.

Notes

Note that for higher dimensional inserts obj=0 behaves very different from obj=[0] just like arr[:,0,:] = values is different from arr[:,[0],:] = values.

Examples

>>> a = np.array([[1, 1], [2, 2], [3, 3]])
>>> a
array([[1, 1],
       [2, 2],
       [3, 3]])
>>> np.insert(a, 1, 5)
array([1, 5, 1, 2, 2, 3, 3])
>>> np.insert(a, 1, 5, axis=1)
array([[1, 5, 1],
       [2, 5, 2],
       [3, 5, 3]])

Difference between sequence and scalars:

>>> np.insert(a, [1], [[1],[2],[3]], axis=1)
array([[1, 1, 1],
       [2, 2, 2],
       [3, 3, 3]])
>>> np.array_equal(np.insert(a, 1, [1, 2, 3], axis=1),
...                np.insert(a, [1], [[1],[2],[3]], axis=1))
True

>>> b = a.flatten()
>>> b
array([1, 1, 2, 2, 3, 3])
>>> np.insert(b, [2, 2], [5, 6])
array([1, 1, 5, 6, 2, 2, 3, 3])

>>> np.insert(b, slice(2, 4), [5, 6])
array([1, 1, 5, 2, 6, 2, 3, 3])

>>> np.insert(b, [2, 2], [7.13, False]) # type casting
array([1, 1, 7, 0, 2, 2, 3, 3])

>>> x = np.arange(8).reshape(2, 4)
>>> idx = (1, 3)
>>> np.insert(x, idx, 999, axis=1)
array([[  0, 999,   1,   2, 999,   3],
       [  4, 999,   5,   6, 999,   7]])
dask.array.isclose(a, b, rtol=1e-05, atol=1e-08, equal_nan=False)

Returns a boolean array where two arrays are element-wise equal within a tolerance.

The tolerance values are positive, typically very small numbers. The relative difference (rtol * abs(b)) and the absolute difference atol are added together to compare against the absolute difference between a and b.

Parameters:

a, b : array_like

Input arrays to compare.

rtol : float

The relative tolerance parameter (see Notes).

atol : float

The absolute tolerance parameter (see Notes).

equal_nan : bool

Whether to compare NaN’s as equal. If True, NaN’s in a will be considered equal to NaN’s in b in the output array.
Returns:

y : array_like

Returns a boolean array of where a and b are equal within the given tolerance. If both a and b are scalars, returns a single boolean value.

See also

allclose

Notes

New in version 1.7.0.

For finite values, isclose uses the following equation to test whether two floating point values are equivalent.

absolute(a - b) <= (atol + rtol * absolute(b))

The above equation is not symmetric in a and b, so that isclose(a, b) might be different from isclose(b, a) in some rare cases.

Examples

>>> np.isclose([1e10,1e-7], [1.00001e10,1e-8])
array([True, False])
>>> np.isclose([1e10,1e-8], [1.00001e10,1e-9])
array([True, True])
>>> np.isclose([1e10,1e-8], [1.0001e10,1e-9])
array([False, True])
>>> np.isclose([1.0, np.nan], [1.0, np.nan])
array([True, False])
>>> np.isclose([1.0, np.nan], [1.0, np.nan], equal_nan=True)
array([True, True])
dask.array.iscomplex(*args, **kwargs)

Returns a bool array, where True if input element is complex.

What is tested is whether the input has a non-zero imaginary part, not if the input type is complex.

Parameters:

x : array_like

Input array.
Returns:

out : ndarray of bools

Output array.

See also

isreal

iscomplexobj
Return True if x is a complex type or an array of complex numbers.

Examples

>>> np.iscomplex([1+1j, 1+0j, 4.5, 3, 2, 2j])  
array([ True, False, False, False, False,  True], dtype=bool)
dask.array.isfinite(x[, out])

Test element-wise for finiteness (not infinity or not Not a Number).

The result is returned as a boolean array.

Parameters:

x : array_like

Input values.

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

y : ndarray, bool

For scalar input, the result is a new boolean with value True if the input is finite; otherwise the value is False (input is either positive infinity, negative infinity or Not a Number).

For array input, the result is a boolean array with the same dimensions as the input and the values are True if the corresponding element of the input is finite; otherwise the values are False (element is either positive infinity, negative infinity or Not a Number).

See also

isinf, isneginf, isposinf, isnan

Notes

Not a Number, positive infinity and negative infinity are considered to be non-finite.

NumPy uses the IEEE Standard for Binary Floating-Point for Arithmetic (IEEE 754). This means that Not a Number is not equivalent to infinity. Also that positive infinity is not equivalent to negative infinity. But infinity is equivalent to positive infinity. Errors result if the second argument is also supplied when x is a scalar input, or if first and second arguments have different shapes.

Examples

>>> np.isfinite(1)  
True
>>> np.isfinite(0)  
True
>>> np.isfinite(np.nan)  
False
>>> np.isfinite(np.inf)  
False
>>> np.isfinite(np.NINF)  
False
>>> np.isfinite([np.log(-1.),1.,np.log(0)])  
array([False,  True, False], dtype=bool)

>>> x = np.array([-np.inf, 0., np.inf])  
>>> y = np.array([2, 2, 2])  
>>> np.isfinite(x, y)  
array([0, 1, 0])
>>> y  
array([0, 1, 0])
dask.array.isinf(x[, out])

Test element-wise for positive or negative infinity.

Returns a boolean array of the same shape as x, True where x == +/-inf, otherwise False.

Parameters:

x : array_like

Input values

out : array_like, optional

An array with the same shape as x to store the result.
Returns:

y : bool (scalar) or boolean ndarray

For scalar input, the result is a new boolean with value True if the input is positive or negative infinity; otherwise the value is False.

For array input, the result is a boolean array with the same shape as the input and the values are True where the corresponding element of the input is positive or negative infinity; elsewhere the values are False. If a second argument was supplied the result is stored there. If the type of that array is a numeric type the result is represented as zeros and ones, if the type is boolean then as False and True, respectively. The return value y is then a reference to that array.

See also

isneginf, isposinf, isnan, isfinite

Notes

NumPy uses the IEEE Standard for Binary Floating-Point for Arithmetic (IEEE 754).

Errors result if the second argument is supplied when the first argument is a scalar, or if the first and second arguments have different shapes.

Examples

>>> np.isinf(np.inf)  
True
>>> np.isinf(np.nan)  
False
>>> np.isinf(np.NINF)  
True
>>> np.isinf([np.inf, -np.inf, 1.0, np.nan])  
array([ True,  True, False, False], dtype=bool)

>>> x = np.array([-np.inf, 0., np.inf])  
>>> y = np.array([2, 2, 2])  
>>> np.isinf(x, y)  
array([1, 0, 1])
>>> y  
array([1, 0, 1])
dask.array.isnan(x[, out])

Test element-wise for NaN and return result as a boolean array.

Parameters:

x : array_like

Input array.
Returns:

y : ndarray or bool

For scalar input, the result is a new boolean with value True if the input is NaN; otherwise the value is False.

For array input, the result is a boolean array of the same dimensions as the input and the values are True if the corresponding element of the input is NaN; otherwise the values are False.

See also

isinf, isneginf, isposinf, isfinite

Notes

NumPy uses the IEEE Standard for Binary Floating-Point for Arithmetic (IEEE 754). This means that Not a Number is not equivalent to infinity.

Examples

>>> np.isnan(np.nan)  
True
>>> np.isnan(np.inf)  
False
>>> np.isnan([np.log(-1.),1.,np.log(0)])  
array([ True, False, False], dtype=bool)
dask.array.isnull(values)

pandas.isnull for dask arrays

dask.array.isreal(*args, **kwargs)

Returns a bool array, where True if input element is real.

If element has complex type with zero complex part, the return value for that element is True.

Parameters:

x : array_like

Input array.
Returns:

out : ndarray, bool

Boolean array of same shape as x.

See also

iscomplex

isrealobj
Return True if x is not a complex type.

Examples

>>> np.isreal([1+1j, 1+0j, 4.5, 3, 2, 2j])  
array([False,  True,  True,  True,  True, False], dtype=bool)
dask.array.ldexp(x1, x2[, out])

Returns x1 * 2**x2, element-wise.

The mantissas x1 and twos exponents x2 are used to construct floating point numbers x1 * 2**x2.

Parameters:

x1 : array_like

Array of multipliers.

x2 : array_like, int

Array of twos exponents.

out : ndarray, optional

Output array for the result.
Returns:

y : ndarray or scalar

The result of x1 * 2**x2.

See also

frexp
Return (y1, y2) from x = y1 * 2**y2, inverse to ldexp.

Notes

Complex dtypes are not supported, they will raise a TypeError.

ldexp is useful as the inverse of frexp, if used by itself it is more clear to simply use the expression x1 * 2**x2.

Examples

>>> np.ldexp(5, np.arange(4))  
array([  5.,  10.,  20.,  40.], dtype=float32)

>>> x = np.arange(6)  
>>> np.ldexp(*np.frexp(x))  
array([ 0.,  1.,  2.,  3.,  4.,  5.])
dask.array.linspace(start, stop, num=50, chunks=None, dtype=None)

Return num evenly spaced values over the closed interval [start, stop].

TODO: implement the endpoint, restep, and dtype keyword args

Parameters:

start : scalar

The starting value of the sequence.

stop : scalar

The last value of the sequence.

num : int, optional

Number of samples to include in the returned dask array, including the endpoints.

chunks : int

The number of samples on each block. Note that the last block will have fewer samples if num % blocksize != 0
Returns:

samples : dask array

See also

dask.array.arange

dask.array.log(x[, out])

Natural logarithm, element-wise.

The natural logarithm log is the inverse of the exponential function, so that log(exp(x)) = x. The natural logarithm is logarithm in base e.

Parameters:

x : array_like

Input value.
Returns:

y : ndarray

The natural logarithm of x, element-wise.

See also

log10, log2, log1p, emath.log

Notes

Logarithm is a multivalued function: for each x there is an infinite number of z such that exp(z) = x. The convention is to return the z whose imaginary part lies in [-pi, pi].

For real-valued input data types, log always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, log is a complex analytical function that has a branch cut [-inf, 0] and is continuous from above on it. log handles the floating-point negative zero as an infinitesimal negative number, conforming to the C99 standard.

References

[R229230]M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 67. http://www.math.sfu.ca/~cbm/aands/
[R230230]Wikipedia, “Logarithm”. http://en.wikipedia.org/wiki/Logarithm

Examples

>>> np.log([1, np.e, np.e**2, 0])  
array([  0.,   1.,   2., -Inf])
dask.array.log10(x[, out])

Return the base 10 logarithm of the input array, element-wise.

Parameters:

x : array_like

Input values.
Returns:

y : ndarray

The logarithm to the base 10 of x, element-wise. NaNs are returned where x is negative.

See also

emath.log10

Notes

Logarithm is a multivalued function: for each x there is an infinite number of z such that 10**z = x. The convention is to return the z whose imaginary part lies in [-pi, pi].

For real-valued input data types, log10 always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, log10 is a complex analytical function that has a branch cut [-inf, 0] and is continuous from above on it. log10 handles the floating-point negative zero as an infinitesimal negative number, conforming to the C99 standard.

References

[R233234]M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 67. http://www.math.sfu.ca/~cbm/aands/
[R234234]Wikipedia, “Logarithm”. http://en.wikipedia.org/wiki/Logarithm

Examples

>>> np.log10([1e-15, -3.])  
array([-15.,  NaN])
dask.array.log1p(x[, out])

Return the natural logarithm of one plus the input array, element-wise.

Calculates log(1 + x).

Parameters:

x : array_like

Input values.
Returns:

y : ndarray

Natural logarithm of 1 + x, element-wise.

See also

expm1
exp(x) - 1, the inverse of log1p.

Notes

For real-valued input, log1p is accurate also for x so small that 1 + x == 1 in floating-point accuracy.

Logarithm is a multivalued function: for each x there is an infinite number of z such that exp(z) = 1 + x. The convention is to return the z whose imaginary part lies in [-pi, pi].

For real-valued input data types, log1p always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, log1p is a complex analytical function that has a branch cut [-inf, -1] and is continuous from above on it. log1p handles the floating-point negative zero as an infinitesimal negative number, conforming to the C99 standard.

References

[R237238]M. Abramowitz and I.A. Stegun, “Handbook of Mathematical Functions”, 10th printing, 1964, pp. 67. http://www.math.sfu.ca/~cbm/aands/
[R238238]Wikipedia, “Logarithm”. http://en.wikipedia.org/wiki/Logarithm

Examples

>>> np.log1p(1e-99)  
1e-99
>>> np.log(1 + 1e-99)  
0.0
dask.array.log2(x[, out])

Base-2 logarithm of x.

Parameters:

x : array_like

Input values.
Returns:

y : ndarray

Base-2 logarithm of x.

See also

log, log10, log1p, emath.log2

Notes

New in version 1.3.0.

Logarithm is a multivalued function: for each x there is an infinite number of z such that 2**z = x. The convention is to return the z whose imaginary part lies in [-pi, pi].

For real-valued input data types, log2 always returns real output. For each value that cannot be expressed as a real number or infinity, it yields nan and sets the invalid floating point error flag.

For complex-valued input, log2 is a complex analytical function that has a branch cut [-inf, 0] and is continuous from above on it. log2 handles the floating-point negative zero as an infinitesimal negative number, conforming to the C99 standard.

Examples

>>> x = np.array([0, 1, 2, 2**4])  
>>> np.log2(x)  
array([-Inf,   0.,   1.,   4.])

>>> xi = np.array([0+1.j, 1, 2+0.j, 4.j])  
>>> np.log2(xi)  
array([ 0.+2.26618007j,  0.+0.j        ,  1.+0.j        ,  2.+2.26618007j])
dask.array.logaddexp(x1, x2[, out])

Logarithm of the sum of exponentiations of the inputs.

Calculates log(exp(x1) + exp(x2)). This function is useful in statistics where the calculated probabilities of events may be so small as to exceed the range of normal floating point numbers. In such cases the logarithm of the calculated probability is stored. This function allows adding probabilities stored in such a fashion.

Parameters:

x1, x2 : array_like

Input values.
Returns:

result : ndarray

Logarithm of exp(x1) + exp(x2).

See also

logaddexp2
Logarithm of the sum of exponentiations of inputs in base 2.

Notes

New in version 1.3.0.

Examples

>>> prob1 = np.log(1e-50)  
>>> prob2 = np.log(2.5e-50)  
>>> prob12 = np.logaddexp(prob1, prob2)  
>>> prob12  
-113.87649168120691
>>> np.exp(prob12)  
3.5000000000000057e-50
dask.array.logaddexp2(x1, x2[, out])

Logarithm of the sum of exponentiations of the inputs in base-2.

Calculates log2(2**x1 + 2**x2). This function is useful in machine learning when the calculated probabilities of events may be so small as to exceed the range of normal floating point numbers. In such cases the base-2 logarithm of the calculated probability can be used instead. This function allows adding probabilities stored in such a fashion.

Parameters:

x1, x2 : array_like

Input values.

out : ndarray, optional

Array to store results in.
Returns:

result : ndarray

Base-2 logarithm of 2**x1 + 2**x2.

See also

logaddexp
Logarithm of the sum of exponentiations of the inputs.

Notes

New in version 1.3.0.

Examples

>>> prob1 = np.log2(1e-50)  
>>> prob2 = np.log2(2.5e-50)  
>>> prob12 = np.logaddexp2(prob1, prob2)  
>>> prob1, prob2, prob12  
(-166.09640474436813, -164.77447664948076, -164.28904982231052)
>>> 2**prob12  
3.4999999999999914e-50
dask.array.logical_and(x1, x2[, out])

Compute the truth value of x1 AND x2 element-wise.

Parameters:

x1, x2 : array_like

Input arrays. x1 and x2 must be of the same shape.
Returns:

y : ndarray or bool

Boolean result with the same shape as x1 and x2 of the logical AND operation on corresponding elements of x1 and x2.

See also

logical_or, logical_not, logical_xor, bitwise_and

Examples

>>> np.logical_and(True, False)  
False
>>> np.logical_and([True, False], [False, False])  
array([False, False], dtype=bool)

>>> x = np.arange(5)  
>>> np.logical_and(x>1, x<4)  
array([False, False,  True,  True, False], dtype=bool)
dask.array.logical_not(x[, out])

Compute the truth value of NOT x element-wise.

Parameters:

x : array_like

Logical NOT is applied to the elements of x.
Returns:

y : bool or ndarray of bool

Boolean result with the same shape as x of the NOT operation on elements of x.

See also

logical_and, logical_or, logical_xor

Examples

>>> np.logical_not(3)  
False
>>> np.logical_not([True, False, 0, 1])  
array([False,  True,  True, False], dtype=bool)

>>> x = np.arange(5)  
>>> np.logical_not(x<3)  
array([False, False, False,  True,  True], dtype=bool)
dask.array.logical_or(x1, x2[, out])

Compute the truth value of x1 OR x2 element-wise.

Parameters:

x1, x2 : array_like

Logical OR is applied to the elements of x1 and x2. They have to be of the same shape.
Returns:

y : ndarray or bool

Boolean result with the same shape as x1 and x2 of the logical OR operation on elements of x1 and x2.

See also

logical_and, logical_not, logical_xor, bitwise_or

Examples

>>> np.logical_or(True, False)  
True
>>> np.logical_or([True, False], [False, False])  
array([ True, False], dtype=bool)

>>> x = np.arange(5)  
>>> np.logical_or(x < 1, x > 3)  
array([ True, False, False, False,  True], dtype=bool)
dask.array.logical_xor(x1, x2[, out])

Compute the truth value of x1 XOR x2, element-wise.

Parameters:

x1, x2 : array_like

Logical XOR is applied to the elements of x1 and x2. They must be broadcastable to the same shape.
Returns:

y : bool or ndarray of bool

Boolean result of the logical XOR operation applied to the elements of x1 and x2; the shape is determined by whether or not broadcasting of one or both arrays was required.

See also

logical_and, logical_or, logical_not, bitwise_xor

Examples

>>> np.logical_xor(True, False)  
True
>>> np.logical_xor([True, True, False, False], [True, False, True, False])  
array([False,  True,  True, False], dtype=bool)

>>> x = np.arange(5)  
>>> np.logical_xor(x < 1, x > 3)  
array([ True, False, False, False,  True], dtype=bool)

Simple example showing support of broadcasting

>>> np.logical_xor(0, np.eye(2))  
array([[ True, False],
       [False,  True]], dtype=bool)
dask.array.max(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the maximum of an array or maximum along an axis.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

Axis or axes along which to operate. By default, flattened input is used.

New in version 1.7.0.

If this is a tuple of ints, the maximum is selected over multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternative output array in which to place the result. Must be of the same shape and buffer length as the expected output. See doc.ufuncs (Section “Output arguments”) for more details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the amax method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

amax : ndarray or scalar

Maximum of a. If axis is None, the result is a scalar value. If axis is given, the result is an array of dimension a.ndim - 1.

See also

amin
The minimum value of an array along a given axis, propagating any NaNs.
nanmax
The maximum value of an array along a given axis, ignoring any NaNs.
maximum
Element-wise maximum of two arrays, propagating any NaNs.
fmax
Element-wise maximum of two arrays, ignoring any NaNs.
argmax
Return the indices of the maximum values.

nanmin, minimum, fmin

Notes

NaN values are propagated, that is if at least one item is NaN, the corresponding max value will be NaN as well. To ignore NaN values (MATLAB behavior), please use nanmax.

Don’t use amax for element-wise comparison of 2 arrays; when a.shape[0] is 2, maximum(a[0], a[1]) is faster than amax(a, axis=0).

Examples

>>> a = np.arange(4).reshape((2,2))
>>> a
array([[0, 1],
       [2, 3]])
>>> np.amax(a)           # Maximum of the flattened array
3
>>> np.amax(a, axis=0)   # Maxima along the first axis
array([2, 3])
>>> np.amax(a, axis=1)   # Maxima along the second axis
array([1, 3])

>>> b = np.arange(5, dtype=np.float)
>>> b[2] = np.NaN
>>> np.amax(b)
nan
>>> np.nanmax(b)
4.0
dask.array.maximum(x1, x2[, out])

Element-wise maximum of array elements.

Compare two arrays and returns a new array containing the element-wise maxima. If one of the elements being compared is a NaN, then that element is returned. If both elements are NaNs then the first is returned. The latter distinction is important for complex NaNs, which are defined as at least one of the real or imaginary parts being a NaN. The net effect is that NaNs are propagated.

Parameters:

x1, x2 : array_like

The arrays holding the elements to be compared. They must have the same shape, or shapes that can be broadcast to a single shape.
Returns:

y : ndarray or scalar

The maximum of x1 and x2, element-wise. Returns scalar if both x1 and x2 are scalars.

See also

minimum
Element-wise minimum of two arrays, propagates NaNs.
fmax
Element-wise maximum of two arrays, ignores NaNs.
amax
The maximum value of an array along a given axis, propagates NaNs.
nanmax
The maximum value of an array along a given axis, ignores NaNs.

fmin, amin, nanmin

Notes

The maximum is equivalent to np.where(x1 >= x2, x1, x2) when neither x1 nor x2 are nans, but it is faster and does proper broadcasting.

Examples

>>> np.maximum([2, 3, 4], [1, 5, 2])  
array([2, 5, 4])

>>> np.maximum(np.eye(2), [0.5, 2]) # broadcasting  
array([[ 1. ,  2. ],
       [ 0.5,  2. ]])

>>> np.maximum([np.nan, 0, np.nan], [0, np.nan, np.nan])  
array([ NaN,  NaN,  NaN])
>>> np.maximum(np.Inf, 1)  
inf
dask.array.mean(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Compute the arithmetic mean along the specified axis.

Returns the average of the array elements. The average is taken over the flattened array by default, otherwise over the specified axis. float64 intermediate and return values are used for integer inputs.

Parameters:

a : array_like

Array containing numbers whose mean is desired. If a is not an array, a conversion is attempted.

axis : None or int or tuple of ints, optional

Axis or axes along which the means are computed. The default is to compute the mean of the flattened array.

New in version 1.7.0.

If this is a tuple of ints, a mean is performed over multiple axes, instead of a single axis or all the axes as before.

dtype : data-type, optional

Type to use in computing the mean. For integer inputs, the default is float64; for floating point inputs, it is the same as the input dtype.

out : ndarray, optional

Alternate output array in which to place the result. The default is None; if provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the mean method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

m : ndarray, see dtype parameter above

If out=None, returns a new array containing the mean values, otherwise a reference to the output array is returned.

See also

average
Weighted average

std, var, nanmean, nanstd, nanvar

Notes

The arithmetic mean is the sum of the elements along the axis divided by the number of elements.

Note that for floating-point input, the mean is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-precision accumulator using the dtype keyword can alleviate this issue.

By default, float16 results are computed using float32 intermediates for extra precision.

Examples

>>> a = np.array([[1, 2], [3, 4]])
>>> np.mean(a)
2.5
>>> np.mean(a, axis=0)
array([ 2.,  3.])
>>> np.mean(a, axis=1)
array([ 1.5,  3.5])

In single precision, mean can be inaccurate:

>>> a = np.zeros((2, 512*512), dtype=np.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.mean(a)
0.54999924

Computing the mean in float64 is more accurate:

>>> np.mean(a, dtype=np.float64)
0.55000000074505806
dask.array.min(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the minimum of an array or minimum along an axis.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

Axis or axes along which to operate. By default, flattened input is used.

New in version 1.7.0.

If this is a tuple of ints, the minimum is selected over multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternative output array in which to place the result. Must be of the same shape and buffer length as the expected output. See doc.ufuncs (Section “Output arguments”) for more details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the amin method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

amin : ndarray or scalar

Minimum of a. If axis is None, the result is a scalar value. If axis is given, the result is an array of dimension a.ndim - 1.

See also

amax
The maximum value of an array along a given axis, propagating any NaNs.
nanmin
The minimum value of an array along a given axis, ignoring any NaNs.
minimum
Element-wise minimum of two arrays, propagating any NaNs.
fmin
Element-wise minimum of two arrays, ignoring any NaNs.
argmin
Return the indices of the minimum values.

nanmax, maximum, fmax

Notes

NaN values are propagated, that is if at least one item is NaN, the corresponding min value will be NaN as well. To ignore NaN values (MATLAB behavior), please use nanmin.

Don’t use amin for element-wise comparison of 2 arrays; when a.shape[0] is 2, minimum(a[0], a[1]) is faster than amin(a, axis=0).

Examples

>>> a = np.arange(4).reshape((2,2))
>>> a
array([[0, 1],
       [2, 3]])
>>> np.amin(a)           # Minimum of the flattened array
0
>>> np.amin(a, axis=0)   # Minima along the first axis
array([0, 1])
>>> np.amin(a, axis=1)   # Minima along the second axis
array([0, 2])

>>> b = np.arange(5, dtype=np.float)
>>> b[2] = np.NaN
>>> np.amin(b)
nan
>>> np.nanmin(b)
0.0
dask.array.minimum(x1, x2[, out])

Element-wise minimum of array elements.

Compare two arrays and returns a new array containing the element-wise minima. If one of the elements being compared is a NaN, then that element is returned. If both elements are NaNs then the first is returned. The latter distinction is important for complex NaNs, which are defined as at least one of the real or imaginary parts being a NaN. The net effect is that NaNs are propagated.

Parameters:

x1, x2 : array_like

The arrays holding the elements to be compared. They must have the same shape, or shapes that can be broadcast to a single shape.
Returns:

y : ndarray or scalar

The minimum of x1 and x2, element-wise. Returns scalar if both x1 and x2 are scalars.

See also

maximum
Element-wise maximum of two arrays, propagates NaNs.
fmin
Element-wise minimum of two arrays, ignores NaNs.
amin
The minimum value of an array along a given axis, propagates NaNs.
nanmin
The minimum value of an array along a given axis, ignores NaNs.

fmax, amax, nanmax

Notes

The minimum is equivalent to np.where(x1 <= x2, x1, x2) when neither x1 nor x2 are NaNs, but it is faster and does proper broadcasting.

Examples

>>> np.minimum([2, 3, 4], [1, 5, 2])  
array([1, 3, 2])

>>> np.minimum(np.eye(2), [0.5, 2]) # broadcasting  
array([[ 0.5,  0. ],
       [ 0. ,  1. ]])

>>> np.minimum([np.nan, 0, np.nan],[0, np.nan, np.nan])  
array([ NaN,  NaN,  NaN])
>>> np.minimum(-np.Inf, 1)  
-inf
dask.array.modf(x[, out1, out2])

Return the fractional and integral parts of an array, element-wise.

The fractional and integral parts are negative if the given number is negative.

Parameters:

x : array_like

Input array.
Returns:

y1 : ndarray

Fractional part of x.

y2 : ndarray

Integral part of x.

Notes

For integer input the return values are floats.

Examples

>>> np.modf([0, 3.5])  
(array([ 0. ,  0.5]), array([ 0.,  3.]))
>>> np.modf(-0.5)  
(-0.5, -0)
dask.array.moment(a, order, axis=None, dtype=None, keepdims=False, ddof=0, split_every=None, out=None)
dask.array.nanargmax(x, axis, **kwargs)
dask.array.nanargmin(x, axis, **kwargs)
dask.array.nancumprod(a, axis=None, dtype=None, out=None)

Return the cumulative product of array elements over a given axis treating Not a Numbers (NaNs) as one. The cumulative product does not change when NaNs are encountered and leading NaNs are replaced by ones.

Ones are returned for slices that are all-NaN or empty.

New in version 1.12.0.

Parameters:

a : array_like

Input array.

axis : int, optional

Axis along which the cumulative product is computed. By default the input is flattened.

dtype : dtype, optional

Type of the returned array, as well as of the accumulator in which the elements are multiplied. If dtype is not specified, it defaults to the dtype of a, unless a has an integer dtype with a precision less than that of the default platform integer. In that case, the default platform integer is used instead.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape and buffer length as the expected output but the type of the resulting values will be cast if necessary.
Returns:

nancumprod : ndarray

A new array holding the result is returned unless out is specified, in which case it is returned.

See also

numpy.cumprod
Cumulative product across array propagating NaNs.
isnan
Show which elements are NaN.

Examples

>>> np.nancumprod(1)
array([1])
>>> np.nancumprod([1])
array([1])
>>> np.nancumprod([1, np.nan])
array([ 1.,  1.])
>>> a = np.array([[1, 2], [3, np.nan]])
>>> np.nancumprod(a)
array([ 1.,  2.,  6.,  6.])
>>> np.nancumprod(a, axis=0)
array([[ 1.,  2.],
       [ 3.,  2.]])
>>> np.nancumprod(a, axis=1)
array([[ 1.,  2.],
       [ 3.,  3.]])
dask.array.nancumsum(a, axis=None, dtype=None, out=None)

Return the cumulative sum of array elements over a given axis treating Not a Numbers (NaNs) as zero. The cumulative sum does not change when NaNs are encountered and leading NaNs are replaced by zeros.

Zeros are returned for slices that are all-NaN or empty.

New in version 1.12.0.

Parameters:

a : array_like

Input array.

axis : int, optional

Axis along which the cumulative sum is computed. The default (None) is to compute the cumsum over the flattened array.

dtype : dtype, optional

Type of the returned array and of the accumulator in which the elements are summed. If dtype is not specified, it defaults to the dtype of a, unless a has an integer dtype with a precision less than that of the default platform integer. In that case, the default platform integer is used.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape and buffer length as the expected output but the type will be cast if necessary. See doc.ufuncs (Section “Output arguments”) for more details.
Returns:

nancumsum : ndarray.

A new array holding the result is returned unless out is specified, in which it is returned. The result has the same size as a, and the same shape as a if axis is not None or a is a 1-d array.

See also

numpy.cumsum
Cumulative sum across array propagating NaNs.
isnan
Show which elements are NaN.

Examples

>>> np.nancumsum(1)
array([1])
>>> np.nancumsum([1])
array([1])
>>> np.nancumsum([1, np.nan])
array([ 1.,  1.])
>>> a = np.array([[1, 2], [3, np.nan]])
>>> np.nancumsum(a)
array([ 1.,  3.,  6.,  6.])
>>> np.nancumsum(a, axis=0)
array([[ 1.,  2.],
       [ 4.,  2.]])
>>> np.nancumsum(a, axis=1)
array([[ 1.,  3.],
       [ 3.,  3.]])
dask.array.nanmax(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the maximum of an array or maximum along an axis, ignoring any NaNs. When all-NaN slices are encountered a RuntimeWarning is raised and NaN is returned for that slice.

Parameters:

a : array_like

Array containing numbers whose maximum is desired. If a is not an array, a conversion is attempted.

axis : int, optional

Axis along which the maximum is computed. The default is to compute the maximum of the flattened array.

out : ndarray, optional

Alternate output array in which to place the result. The default is None; if provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details.

New in version 1.8.0.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original a.

If the value is anything but the default, then keepdims will be passed through to the max method of sub-classes of ndarray. If the sub-classes methods does not implement keepdims any exceptions will be raised.

New in version 1.8.0.
Returns:

nanmax : ndarray

An array with the same shape as a, with the specified axis removed. If a is a 0-d array, or if axis is None, an ndarray scalar is returned. The same dtype as a is returned.

See also

nanmin
The minimum value of an array along a given axis, ignoring any NaNs.
amax
The maximum value of an array along a given axis, propagating any NaNs.
fmax
Element-wise maximum of two arrays, ignoring any NaNs.
maximum
Element-wise maximum of two arrays, propagating any NaNs.
isnan
Shows which elements are Not a Number (NaN).
isfinite
Shows which elements are neither NaN nor infinity.

amin, fmin, minimum

Notes

NumPy uses the IEEE Standard for Binary Floating-Point for Arithmetic (IEEE 754). This means that Not a Number is not equivalent to infinity. Positive infinity is treated as a very large number and negative infinity is treated as a very small (i.e. negative) number.

If the input has a integer type the function is equivalent to np.max.

Examples

>>> a = np.array([[1, 2], [3, np.nan]])
>>> np.nanmax(a)
3.0
>>> np.nanmax(a, axis=0)
array([ 3.,  2.])
>>> np.nanmax(a, axis=1)
array([ 2.,  3.])

When positive infinity and negative infinity are present:

>>> np.nanmax([1, 2, np.nan, np.NINF])
2.0
>>> np.nanmax([1, 2, np.nan, np.inf])
inf
dask.array.nanmean(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Compute the arithmetic mean along the specified axis, ignoring NaNs.

Returns the average of the array elements. The average is taken over the flattened array by default, otherwise over the specified axis. float64 intermediate and return values are used for integer inputs.

For all-NaN slices, NaN is returned and a RuntimeWarning is raised.

New in version 1.8.0.

Parameters:

a : array_like

Array containing numbers whose mean is desired. If a is not an array, a conversion is attempted.

axis : int, optional

Axis along which the means are computed. The default is to compute the mean of the flattened array.

dtype : data-type, optional

Type to use in computing the mean. For integer inputs, the default is float64; for inexact inputs, it is the same as the input dtype.

out : ndarray, optional

Alternate output array in which to place the result. The default is None; if provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original a.

If the value is anything but the default, then keepdims will be passed through to the mean or sum methods of sub-classes of ndarray. If the sub-classes methods does not implement keepdims any exceptions will be raised.
Returns:

m : ndarray, see dtype parameter above

If out=None, returns a new array containing the mean values, otherwise a reference to the output array is returned. Nan is returned for slices that contain only NaNs.

See also

average
Weighted average
mean
Arithmetic mean taken while not ignoring NaNs

var, nanvar

Notes

The arithmetic mean is the sum of the non-NaN elements along the axis divided by the number of non-NaN elements.

Note that for floating-point input, the mean is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32. Specifying a higher-precision accumulator using the dtype keyword can alleviate this issue.

Examples

>>> a = np.array([[1, np.nan], [3, 4]])
>>> np.nanmean(a)
2.6666666666666665
>>> np.nanmean(a, axis=0)
array([ 2.,  4.])
>>> np.nanmean(a, axis=1)
array([ 1.,  3.5])
dask.array.nanmin(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return minimum of an array or minimum along an axis, ignoring any NaNs. When all-NaN slices are encountered a RuntimeWarning is raised and Nan is returned for that slice.

Parameters:

a : array_like

Array containing numbers whose minimum is desired. If a is not an array, a conversion is attempted.

axis : int, optional

Axis along which the minimum is computed. The default is to compute the minimum of the flattened array.

out : ndarray, optional

Alternate output array in which to place the result. The default is None; if provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details.

New in version 1.8.0.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original a.

If the value is anything but the default, then keepdims will be passed through to the min method of sub-classes of ndarray. If the sub-classes methods does not implement keepdims any exceptions will be raised.

New in version 1.8.0.
Returns:

nanmin : ndarray

An array with the same shape as a, with the specified axis removed. If a is a 0-d array, or if axis is None, an ndarray scalar is returned. The same dtype as a is returned.

See also

nanmax
The maximum value of an array along a given axis, ignoring any NaNs.
amin
The minimum value of an array along a given axis, propagating any NaNs.
fmin
Element-wise minimum of two arrays, ignoring any NaNs.
minimum
Element-wise minimum of two arrays, propagating any NaNs.
isnan
Shows which elements are Not a Number (NaN).
isfinite
Shows which elements are neither NaN nor infinity.

amax, fmax, maximum

Notes

NumPy uses the IEEE Standard for Binary Floating-Point for Arithmetic (IEEE 754). This means that Not a Number is not equivalent to infinity. Positive infinity is treated as a very large number and negative infinity is treated as a very small (i.e. negative) number.

If the input has a integer type the function is equivalent to np.min.

Examples

>>> a = np.array([[1, 2], [3, np.nan]])
>>> np.nanmin(a)
1.0
>>> np.nanmin(a, axis=0)
array([ 1.,  2.])
>>> np.nanmin(a, axis=1)
array([ 1.,  3.])

When positive infinity and negative infinity are present:

>>> np.nanmin([1, 2, np.nan, np.inf])
1.0
>>> np.nanmin([1, 2, np.nan, np.NINF])
-inf
dask.array.nanprod(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the product of array elements over a given axis treating Not a Numbers (NaNs) as ones.

One is returned for slices that are all-NaN or empty.

New in version 1.10.0.

Parameters:

a : array_like

Array containing numbers whose sum is desired. If a is not an array, a conversion is attempted.

axis : int, optional

Axis along which the product is computed. The default is to compute the product of the flattened array.

dtype : data-type, optional

The type of the returned array and of the accumulator in which the elements are summed. By default, the dtype of a is used. An exception is when a has an integer type with less precision than the platform (u)intp. In that case, the default will be either (u)int32 or (u)int64 depending on whether the platform is 32 or 64 bits. For inexact inputs, dtype must be inexact.

out : ndarray, optional

Alternate output array in which to place the result. The default is None. If provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details. The casting of NaN to integer can yield unexpected results.

keepdims : bool, optional

If True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original arr.
Returns:

nanprod : ndarray

A new array holding the result is returned unless out is specified, in which case it is returned.

See also

numpy.prod
Product across array propagating NaNs.
isnan
Show which elements are NaN.

Examples

>>> np.nanprod(1)
1
>>> np.nanprod([1])
1
>>> np.nanprod([1, np.nan])
1.0
>>> a = np.array([[1, 2], [3, np.nan]])
>>> np.nanprod(a)
6.0
>>> np.nanprod(a, axis=0)
array([ 3.,  2.])
dask.array.nanstd(a, axis=None, dtype=None, out=None, ddof=0, keepdims=<class 'numpy._globals._NoValue'>)

Compute the standard deviation along the specified axis, while ignoring NaNs.

Returns the standard deviation, a measure of the spread of a distribution, of the non-NaN array elements. The standard deviation is computed for the flattened array by default, otherwise over the specified axis.

For all-NaN slices or slices with zero degrees of freedom, NaN is returned and a RuntimeWarning is raised.

New in version 1.8.0.

Parameters:

a : array_like

Calculate the standard deviation of the non-NaN values.

axis : int, optional

Axis along which the standard deviation is computed. The default is to compute the standard deviation of the flattened array.

dtype : dtype, optional

Type to use in computing the standard deviation. For arrays of integer type the default is float64, for arrays of float types it is the same as the array type.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output but the type (of the calculated values) will be cast if necessary.

ddof : int, optional

Means Delta Degrees of Freedom. The divisor used in calculations is N - ddof, where N represents the number of non-NaN elements. By default ddof is zero.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original a.

If this value is anything but the default it is passed through as-is to the relevant functions of the sub-classes. If these functions do not have a keepdims kwarg, a RuntimeError will be raised.
Returns:

standard_deviation : ndarray, see dtype parameter above.

If out is None, return a new array containing the standard deviation, otherwise return a reference to the output array. If ddof is >= the number of non-NaN elements in a slice or the slice contains only NaNs, then the result for that slice is NaN.

See also

var, mean, std, nanvar, nanmean

numpy.doc.ufuncs
Section “Output arguments”

Notes

The standard deviation is the square root of the average of the squared deviations from the mean: std = sqrt(mean(abs(x - x.mean())**2)).

The average squared deviation is normally calculated as x.sum() / N, where N = len(x). If, however, ddof is specified, the divisor N - ddof is used instead. In standard statistical practice, ddof=1 provides an unbiased estimator of the variance of the infinite population. ddof=0 provides a maximum likelihood estimate of the variance for normally distributed variables. The standard deviation computed in this function is the square root of the estimated variance, so even with ddof=1, it will not be an unbiased estimate of the standard deviation per se.

Note that, for complex numbers, std takes the absolute value before squaring, so that the result is always real and nonnegative.

For floating-point input, the std is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-accuracy accumulator using the dtype keyword can alleviate this issue.

Examples

>>> a = np.array([[1, np.nan], [3, 4]])
>>> np.nanstd(a)
1.247219128924647
>>> np.nanstd(a, axis=0)
array([ 1.,  0.])
>>> np.nanstd(a, axis=1)
array([ 0.,  0.5])
dask.array.nansum(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the sum of array elements over a given axis treating Not a Numbers (NaNs) as zero.

In NumPy versions <= 1.8.0 Nan is returned for slices that are all-NaN or empty. In later versions zero is returned.

Parameters:

a : array_like

Array containing numbers whose sum is desired. If a is not an array, a conversion is attempted.

axis : int, optional

Axis along which the sum is computed. The default is to compute the sum of the flattened array.

dtype : data-type, optional

The type of the returned array and of the accumulator in which the elements are summed. By default, the dtype of a is used. An exception is when a has an integer type with less precision than the platform (u)intp. In that case, the default will be either (u)int32 or (u)int64 depending on whether the platform is 32 or 64 bits. For inexact inputs, dtype must be inexact.

New in version 1.8.0.

out : ndarray, optional

Alternate output array in which to place the result. The default is None. If provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details. The casting of NaN to integer can yield unexpected results.

New in version 1.8.0.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original a.

If the value is anything but the default, then keepdims will be passed through to the mean or sum methods of sub-classes of ndarray. If the sub-classes methods does not implement keepdims any exceptions will be raised.

New in version 1.8.0.
Returns:

nansum : ndarray.

A new array holding the result is returned unless out is specified, in which it is returned. The result has the same size as a, and the same shape as a if axis is not None or a is a 1-d array.

See also

numpy.sum
Sum across array propagating NaNs.
isnan
Show which elements are NaN.
isfinite
Show which elements are not NaN or +/-inf.

Notes

If both positive and negative infinity are present, the sum will be Not A Number (NaN).

Examples

>>> np.nansum(1)
1
>>> np.nansum([1])
1
>>> np.nansum([1, np.nan])
1.0
>>> a = np.array([[1, 1], [1, np.nan]])
>>> np.nansum(a)
3.0
>>> np.nansum(a, axis=0)
array([ 2.,  1.])
>>> np.nansum([1, np.nan, np.inf])
inf
>>> np.nansum([1, np.nan, np.NINF])
-inf
>>> np.nansum([1, np.nan, np.inf, -np.inf]) # both +/- infinity present
nan
dask.array.nanvar(a, axis=None, dtype=None, out=None, ddof=0, keepdims=<class 'numpy._globals._NoValue'>)

Compute the variance along the specified axis, while ignoring NaNs.

Returns the variance of the array elements, a measure of the spread of a distribution. The variance is computed for the flattened array by default, otherwise over the specified axis.

For all-NaN slices or slices with zero degrees of freedom, NaN is returned and a RuntimeWarning is raised.

New in version 1.8.0.

Parameters:

a : array_like

Array containing numbers whose variance is desired. If a is not an array, a conversion is attempted.

axis : int, optional

Axis along which the variance is computed. The default is to compute the variance of the flattened array.

dtype : data-type, optional

Type to use in computing the variance. For arrays of integer type the default is float32; for arrays of float types it is the same as the array type.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output, but the type is cast if necessary.

ddof : int, optional

“Delta Degrees of Freedom”: the divisor used in the calculation is N - ddof, where N represents the number of non-NaN elements. By default ddof is zero.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original a.
Returns:

variance : ndarray, see dtype parameter above

If out is None, return a new array containing the variance, otherwise return a reference to the output array. If ddof is >= the number of non-NaN elements in a slice or the slice contains only NaNs, then the result for that slice is NaN.

See also

std
Standard deviation
mean
Average
var
Variance while not ignoring NaNs

nanstd, nanmean

numpy.doc.ufuncs
Section “Output arguments”

Notes

The variance is the average of the squared deviations from the mean, i.e., var = mean(abs(x - x.mean())**2).

The mean is normally calculated as x.sum() / N, where N = len(x). If, however, ddof is specified, the divisor N - ddof is used instead. In standard statistical practice, ddof=1 provides an unbiased estimator of the variance of a hypothetical infinite population. ddof=0 provides a maximum likelihood estimate of the variance for normally distributed variables.

Note that for complex numbers, the absolute value is taken before squaring, so that the result is always real and nonnegative.

For floating-point input, the variance is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-accuracy accumulator using the dtype keyword can alleviate this issue.

For this function to work on sub-classes of ndarray, they must define sum with the kwarg keepdims

Examples

>>> a = np.array([[1, np.nan], [3, 4]])
>>> np.var(a)
1.5555555555555554
>>> np.nanvar(a, axis=0)
array([ 1.,  0.])
>>> np.nanvar(a, axis=1)
array([ 0.,  0.25])
dask.array.nextafter(x1, x2[, out])

Return the next floating-point value after x1 towards x2, element-wise.

Parameters:

x1 : array_like

Values to find the next representable value of.

x2 : array_like

The direction where to look for the next representable value of x1.

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

out : array_like

The next representable values of x1 in the direction of x2.

Examples

>>> eps = np.finfo(np.float64).eps  
>>> np.nextafter(1, 2) == eps + 1  
True
>>> np.nextafter([1, 2], [2, 1]) == [eps + 1, 2 - eps]  
array([ True,  True], dtype=bool)
dask.array.notnull(values)

pandas.notnull for dask arrays

dask.array.ones()

Blocked variant of ones

Follows the signature of ones exactly except that it also requires a keyword argument chunks=(...)

Original signature follows below.

Return a new array of given shape and type, filled with ones.

Parameters:

shape : int or sequence of ints

Shape of the new array, e.g., (2, 3) or 2.

dtype : data-type, optional

The desired data-type for the array, e.g., numpy.int8. Default is numpy.float64.

order : {‘C’, ‘F’}, optional

Whether to store multidimensional data in C- or Fortran-contiguous (row- or column-wise) order in memory.
Returns:

out : ndarray

Array of ones with the given shape, dtype, and order.

See also

zeros, ones_like

Examples

>>> np.ones(5)
array([ 1.,  1.,  1.,  1.,  1.])

>>> np.ones((5,), dtype=np.int)
array([1, 1, 1, 1, 1])

>>> np.ones((2, 1))
array([[ 1.],
       [ 1.]])

>>> s = (2,2)
>>> np.ones(s)
array([[ 1.,  1.],
       [ 1.,  1.]])
dask.array.percentile(a, q, interpolation='linear')

Approximate percentile of 1-D array

See numpy.percentile for more information

dask.array.prod(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the product of array elements over a given axis.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

Axis or axes along which a product is performed. The default, axis=None, will calculate the product of all the elements in the input array. If axis is negative it counts from the last to the first axis.

New in version 1.7.0.

If axis is a tuple of ints, a product is performed on all of the axes specified in the tuple instead of a single axis or all the axes as before.

dtype : dtype, optional

The type of the returned array, as well as of the accumulator in which the elements are multiplied. The dtype of a is used by default unless a has an integer dtype of less precision than the default platform integer. In that case, if a is signed then the platform integer is used while if a is unsigned then an unsigned integer of the same precision as the platform integer is used.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the prod method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

product_along_axis : ndarray, see dtype parameter above.

An array shaped as a but with the specified axis removed. Returns a reference to out if specified.

See also

ndarray.prod
equivalent method
numpy.doc.ufuncs
Section “Output arguments”

Notes

Arithmetic is modular when using integer types, and no error is raised on overflow. That means that, on a 32-bit platform:

>>> x = np.array([536870910, 536870910, 536870910, 536870910])
>>> np.prod(x) #random
16

The product of an empty array is the neutral element 1:

>>> np.prod([])
1.0

Examples

By default, calculate the product of all elements:

>>> np.prod([1.,2.])
2.0

Even when the input array is two-dimensional:

>>> np.prod([[1.,2.],[3.,4.]])
24.0

But we can also specify the axis over which to multiply:

>>> np.prod([[1.,2.],[3.,4.]], axis=1)
array([  2.,  12.])

If the type of x is unsigned, then the output type is the unsigned platform integer:

>>> x = np.array([1, 2, 3], dtype=np.uint8)
>>> np.prod(x).dtype == np.uint
True

If x is of a signed integer type, then the output type is the default platform integer:

>>> x = np.array([1, 2, 3], dtype=np.int8)
>>> np.prod(x).dtype == np.int
True
dask.array.rad2deg(x[, out])

Convert angles from radians to degrees.

Parameters:

x : array_like

Angle in radians.

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

y : ndarray

The corresponding angle in degrees.

See also

deg2rad
Convert angles from degrees to radians.
unwrap
Remove large jumps in angle by wrapping.

Notes

New in version 1.3.0.

rad2deg(x) is 180 * x / pi.

Examples

>>> np.rad2deg(np.pi/2)  
90.0
dask.array.radians(x[, out])

Convert angles from degrees to radians.

Parameters:

x : array_like

Input array in degrees.

out : ndarray, optional

Output array of same shape as x.
Returns:

y : ndarray

The corresponding radian values.

See also

deg2rad
equivalent function

Examples

Convert a degree array to radians

>>> deg = np.arange(12.) * 30.  
>>> np.radians(deg)  
array([ 0.        ,  0.52359878,  1.04719755,  1.57079633,  2.0943951 ,
        2.61799388,  3.14159265,  3.66519143,  4.1887902 ,  4.71238898,
        5.23598776,  5.75958653])

>>> out = np.zeros((deg.shape))  
>>> ret = np.radians(deg, out)  
>>> ret is out  
True
dask.array.ravel(a, order='C')

Return a contiguous flattened array.

A 1-D array, containing the elements of the input, is returned. A copy is made only if needed.

As of NumPy 1.10, the returned array will have the same type as the input array. (for example, a masked array will be returned for a masked array input)

Parameters:

a : array_like

Input array. The elements in a are read in the order specified by order, and packed as a 1-D array.

order : {‘C’,’F’, ‘A’, ‘K’}, optional

The elements of a are read using this index order. ‘C’ means to index the elements in row-major, C-style order, with the last axis index changing fastest, back to the first axis index changing slowest. ‘F’ means to index the elements in column-major, Fortran-style order, with the first index changing fastest, and the last index changing slowest. Note that the ‘C’ and ‘F’ options take no account of the memory layout of the underlying array, and only refer to the order of axis indexing. ‘A’ means to read the elements in Fortran-like index order if a is Fortran contiguous in memory, C-like order otherwise. ‘K’ means to read the elements in the order they occur in memory, except for reversing the data when strides are negative. By default, ‘C’ index order is used.
Returns:

y : array_like

If a is a matrix, y is a 1-D ndarray, otherwise y is an array of the same subtype as a. The shape of the returned array is (a.size,). Matrices are special cased for backward compatibility.

See also

ndarray.flat
1-D iterator over an array.
ndarray.flatten
1-D array copy of the elements of an array in row-major order.
ndarray.reshape
Change the shape of an array without changing its data.

Notes

In row-major, C-style order, in two dimensions, the row index varies the slowest, and the column index the quickest. This can be generalized to multiple dimensions, where row-major order implies that the index along the first axis varies slowest, and the index along the last quickest. The opposite holds for column-major, Fortran-style index ordering.

When a view is desired in as many cases as possible, arr.reshape(-1) may be preferable.

Examples

It is equivalent to reshape(-1, order=order).

>>> x = np.array([[1, 2, 3], [4, 5, 6]])
>>> print(np.ravel(x))
[1 2 3 4 5 6]

>>> print(x.reshape(-1))
[1 2 3 4 5 6]

>>> print(np.ravel(x, order='F'))
[1 4 2 5 3 6]

When order is ‘A’, it will preserve the array’s ‘C’ or ‘F’ ordering:

>>> print(np.ravel(x.T))
[1 4 2 5 3 6]
>>> print(np.ravel(x.T, order='A'))
[1 2 3 4 5 6]

When order is ‘K’, it will preserve orderings that are neither ‘C’ nor ‘F’, but won’t reverse axes:

>>> a = np.arange(3)[::-1]; a
array([2, 1, 0])
>>> a.ravel(order='C')
array([2, 1, 0])
>>> a.ravel(order='K')
array([2, 1, 0])

>>> a = np.arange(12).reshape(2,3,2).swapaxes(1,2); a
array([[[ 0,  2,  4],
        [ 1,  3,  5]],
       [[ 6,  8, 10],
        [ 7,  9, 11]]])
>>> a.ravel(order='C')
array([ 0,  2,  4,  1,  3,  5,  6,  8, 10,  7,  9, 11])
>>> a.ravel(order='K')
array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11])
dask.array.real(*args, **kwargs)

Return the real part of the elements of the array.

Parameters:

val : array_like

Input array.
Returns:

out : ndarray

Output array. If val is real, the type of val is used for the output. If val has complex elements, the returned type is float.

See also

real_if_close, imag, angle

Examples

>>> a = np.array([1+2j, 3+4j, 5+6j])  
>>> a.real  
array([ 1.,  3.,  5.])
>>> a.real = 9  
>>> a  
array([ 9.+2.j,  9.+4.j,  9.+6.j])
>>> a.real = np.array([9, 8, 7])  
>>> a  
array([ 9.+2.j,  8.+4.j,  7.+6.j])
dask.array.rechunk(x, chunks, threshold=4, block_size_limit=100000000.0)

Convert blocks in dask array x for new chunks.

>>> import dask.array as da
>>> a = np.random.uniform(0, 1, 7**4).reshape((7,) * 4)
>>> x = da.from_array(a, chunks=((2, 3, 2),)*4)
>>> x.chunks
((2, 3, 2), (2, 3, 2), (2, 3, 2), (2, 3, 2))

>>> y = rechunk(x, chunks=((2, 4, 1), (4, 2, 1), (4, 3), (7,)))
>>> y.chunks
((2, 4, 1), (4, 2, 1), (4, 3), (7,))

chunks also accept dict arguments mapping axis to blockshape

>>> y = rechunk(x, chunks={1: 2})  # rechunk axis 1 with blockshape 2
Parameters:

x: dask array

chunks: tuple

The new block dimensions to create

threshold: int

The graph growth factor under which we don’t bother introducing an intermediate step

block_size_limit: int

The maximum block size (in bytes) we want to produce during an intermediate step
dask.array.repeat(a, repeats, axis=None)

Repeat elements of an array.

Parameters:

a : array_like

Input array.

repeats : int or array of ints

The number of repetitions for each element. repeats is broadcasted to fit the shape of the given axis.

axis : int, optional

The axis along which to repeat values. By default, use the flattened input array, and return a flat output array.
Returns:

repeated_array : ndarray

Output array which has the same shape as a, except along the given axis.

See also

tile
Tile an array.

Examples

>>> np.repeat(3, 4)
array([3, 3, 3, 3])
>>> x = np.array([[1,2],[3,4]])
>>> np.repeat(x, 2)
array([1, 1, 2, 2, 3, 3, 4, 4])
>>> np.repeat(x, 3, axis=1)
array([[1, 1, 1, 2, 2, 2],
       [3, 3, 3, 4, 4, 4]])
>>> np.repeat(x, [1, 2], axis=0)
array([[1, 2],
       [3, 4],
       [3, 4]])
dask.array.reshape(x, shape)

Reshape array to new shape

This is a parallelized version of the np.reshape function with the following limitations:

  1. It assumes that the array is stored in C-order
  2. It only allows for reshapings that collapse or merge dimensions like (1, 2, 3, 4) -> (1, 6, 4) or (64,) -> (4, 4, 4)

When communication is necessary this algorithm depends on the logic within rechunk. It endeavors to keep chunk sizes roughly the same when possible.

See also

dask.array.rechunk, numpy.reshape

dask.array.rint(x[, out])

Round elements of the array to the nearest integer.

Parameters:

x : array_like

Input array.
Returns:

out : ndarray or scalar

Output array is same shape and type as x.

See also

ceil, floor, trunc

Examples

>>> a = np.array([-1.7, -1.5, -0.2, 0.2, 1.5, 1.7, 2.0])  
>>> np.rint(a)  
array([-2., -2., -0.,  0.,  2.,  2.,  2.])
dask.array.roll(a, shift, axis=None)

Roll array elements along a given axis.

Elements that roll beyond the last position are re-introduced at the first.

Parameters:

a : array_like

Input array.

shift : int or tuple of ints

The number of places by which elements are shifted. If a tuple, then axis must be a tuple of the same size, and each of the given axes is shifted by the corresponding number. If an int while axis is a tuple of ints, then the same value is used for all given axes.

axis : int or tuple of ints, optional

Axis or axes along which elements are shifted. By default, the array is flattened before shifting, after which the original shape is restored.
Returns:

res : ndarray

Output array, with the same shape as a.

See also

rollaxis
Roll the specified axis backwards, until it lies in a given position.

Notes

New in version 1.12.0.

Supports rolling over multiple dimensions simultaneously.

Examples

>>> x = np.arange(10)
>>> np.roll(x, 2)
array([8, 9, 0, 1, 2, 3, 4, 5, 6, 7])

>>> x2 = np.reshape(x, (2,5))
>>> x2
array([[0, 1, 2, 3, 4],
       [5, 6, 7, 8, 9]])
>>> np.roll(x2, 1)
array([[9, 0, 1, 2, 3],
       [4, 5, 6, 7, 8]])
>>> np.roll(x2, 1, axis=0)
array([[5, 6, 7, 8, 9],
       [0, 1, 2, 3, 4]])
>>> np.roll(x2, 1, axis=1)
array([[4, 0, 1, 2, 3],
       [9, 5, 6, 7, 8]])
dask.array.round(a, decimals=0, out=None)

Round an array to the given number of decimals.

Refer to around for full documentation.

See also

around
equivalent function
dask.array.sign(x[, out])

Returns an element-wise indication of the sign of a number.

The sign function returns -1 if x < 0, 0 if x==0, 1 if x > 0. nan is returned for nan inputs.

For complex inputs, the sign function returns sign(x.real) + 0j if x.real != 0 else sign(x.imag) + 0j.

complex(nan, 0) is returned for complex nan inputs.

Parameters:

x : array_like

Input values.
Returns:

y : ndarray

The sign of x.

Notes

There is more than one definition of sign in common use for complex numbers. The definition used here is equivalent to \(x/\sqrt{x*x}\) which is different from a common alternative, \(x/|x|\).

Examples

>>> np.sign([-5., 4.5])  
array([-1.,  1.])
>>> np.sign(0)  
0
>>> np.sign(5-2j)  
(1+0j)
dask.array.signbit(x[, out])

Returns element-wise True where signbit is set (less than zero).

Parameters:

x : array_like

The input value(s).

out : ndarray, optional

Array into which the output is placed. Its type is preserved and it must be of the right shape to hold the output. See doc.ufuncs.
Returns:

result : ndarray of bool

Output array, or reference to out if that was supplied.

Examples

>>> np.signbit(-1.2)  
True
>>> np.signbit(np.array([1, -2.3, 2.1]))  
array([False,  True, False], dtype=bool)
dask.array.sin(x[, out])

Trigonometric sine, element-wise.

Parameters:

x : array_like

Angle, in radians (\(2 \pi\) rad equals 360 degrees).
Returns:

y : array_like

The sine of each element of x.

See also

arcsin, sinh, cos

Notes

The sine is one of the fundamental functions of trigonometry (the mathematical study of triangles). Consider a circle of radius 1 centered on the origin. A ray comes in from the \(+x\) axis, makes an angle at the origin (measured counter-clockwise from that axis), and departs from the origin. The \(y\) coordinate of the outgoing ray’s intersection with the unit circle is the sine of that angle. It ranges from -1 for \(x=3\pi / 2\) to +1 for \(\pi / 2.\) The function has zeroes where the angle is a multiple of \(\pi\). Sines of angles between \(\pi\) and \(2\pi\) are negative. The numerous properties of the sine and related functions are included in any standard trigonometry text.

Examples

Print sine of one angle:

>>> np.sin(np.pi/2.)  
1.0

Print sines of an array of angles given in degrees:

>>> np.sin(np.array((0., 30., 45., 60., 90.)) * np.pi / 180. )  
array([ 0.        ,  0.5       ,  0.70710678,  0.8660254 ,  1.        ])

Plot the sine function:

>>> import matplotlib.pylab as plt  
>>> x = np.linspace(-np.pi, np.pi, 201)  
>>> plt.plot(x, np.sin(x))  
>>> plt.xlabel('Angle [rad]')  
>>> plt.ylabel('sin(x)')  
>>> plt.axis('tight')  
>>> plt.show()
dask.array.sinh(x[, out])

Hyperbolic sine, element-wise.

Equivalent to 1/2 * (np.exp(x) - np.exp(-x)) or -1j * np.sin(1j*x).

Parameters:

x : array_like

Input array.

out : ndarray, optional

Output array of same shape as x.
Returns:

y : ndarray

The corresponding hyperbolic sine values.
Raises:

ValueError: invalid return array shape

if out is provided and out.shape != x.shape (See Examples)

Notes

If out is provided, the function writes the result into it, and returns a reference to out. (See Examples)

References

M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions. New York, NY: Dover, 1972, pg. 83.

Examples

>>> np.sinh(0)  
0.0
>>> np.sinh(np.pi*1j/2)  
1j
>>> np.sinh(np.pi*1j) # (exact value is 0)  
1.2246063538223773e-016j
>>> # Discrepancy due to vagaries of floating point arithmetic.

>>> # Example of providing the optional output parameter
>>> out2 = np.sinh([0.1], out1)  
>>> out2 is out1  
True

>>> # Example of ValueError due to provision of shape mis-matched `out`
>>> np.sinh(np.zeros((3,3)),np.zeros((2,2)))  
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: invalid return array shape
dask.array.sqrt(x[, out])

Return the positive square-root of an array, element-wise.

Parameters:

x : array_like

The values whose square-roots are required.

out : ndarray, optional

Alternate array object in which to put the result; if provided, it must have the same shape as x
Returns:

y : ndarray

An array of the same shape as x, containing the positive square-root of each element in x. If any element in x is complex, a complex array is returned (and the square-roots of negative reals are calculated). If all of the elements in x are real, so is y, with negative elements returning nan. If out was provided, y is a reference to it.

See also

lib.scimath.sqrt
A version which returns complex numbers when given negative reals.

Notes

sqrt has–consistent with common convention–as its branch cut the real “interval” [-inf, 0), and is continuous from above on it. A branch cut is a curve in the complex plane across which a given complex function fails to be continuous.

Examples

>>> np.sqrt([1,4,9])  
array([ 1.,  2.,  3.])

>>> np.sqrt([4, -1, -3+4J])  
array([ 2.+0.j,  0.+1.j,  1.+2.j])

>>> np.sqrt([4, -1, numpy.inf])  
array([  2.,  NaN,  Inf])
dask.array.square(x[, out])

Return the element-wise square of the input.

Parameters:

x : array_like

Input data.
Returns:

out : ndarray

Element-wise x*x, of the same shape and dtype as x. Returns scalar if x is a scalar.

See also

numpy.linalg.matrix_power, sqrt, power

Examples

>>> np.square([-1j, 1])  
array([-1.-0.j,  1.+0.j])
dask.array.squeeze(a, axis=None)

Remove single-dimensional entries from the shape of an array.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

New in version 1.7.0.

Selects a subset of the single-dimensional entries in the shape. If an axis is selected with shape entry greater than one, an error is raised.
Returns:

squeezed : ndarray

The input array, but with all or a subset of the dimensions of length 1 removed. This is always a itself or a view into a.

Examples

>>> x = np.array([[[0], [1], [2]]])
>>> x.shape
(1, 3, 1)
>>> np.squeeze(x).shape
(3,)
>>> np.squeeze(x, axis=(2,)).shape
(1, 3)
dask.array.stack(seq, axis=0)

Stack arrays along a new axis

Given a sequence of dask Arrays form a new dask Array by stacking them along a new dimension (axis=0 by default)

See also

concatenate

Examples

Create slices

>>> import dask.array as da
>>> import numpy as np

>>> data = [from_array(np.ones((4, 4)), chunks=(2, 2))
...          for i in range(3)]

>>> x = da.stack(data, axis=0)
>>> x.shape
(3, 4, 4)

>>> da.stack(data, axis=1).shape
(4, 3, 4)

>>> da.stack(data, axis=-1).shape
(4, 4, 3)

Result is a new dask Array

dask.array.std(a, axis=None, dtype=None, out=None, ddof=0, keepdims=<class 'numpy._globals._NoValue'>)

Compute the standard deviation along the specified axis.

Returns the standard deviation, a measure of the spread of a distribution, of the array elements. The standard deviation is computed for the flattened array by default, otherwise over the specified axis.

Parameters:

a : array_like

Calculate the standard deviation of these values.

axis : None or int or tuple of ints, optional

Axis or axes along which the standard deviation is computed. The default is to compute the standard deviation of the flattened array.

New in version 1.7.0.

If this is a tuple of ints, a standard deviation is performed over multiple axes, instead of a single axis or all the axes as before.

dtype : dtype, optional

Type to use in computing the standard deviation. For arrays of integer type the default is float64, for arrays of float types it is the same as the array type.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output but the type (of the calculated values) will be cast if necessary.

ddof : int, optional

Means Delta Degrees of Freedom. The divisor used in calculations is N - ddof, where N represents the number of elements. By default ddof is zero.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the std method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

standard_deviation : ndarray, see dtype parameter above.

If out is None, return a new array containing the standard deviation, otherwise return a reference to the output array.

See also

var, mean, nanmean, nanstd, nanvar

numpy.doc.ufuncs
Section “Output arguments”

Notes

The standard deviation is the square root of the average of the squared deviations from the mean, i.e., std = sqrt(mean(abs(x - x.mean())**2)).

The average squared deviation is normally calculated as x.sum() / N, where N = len(x). If, however, ddof is specified, the divisor N - ddof is used instead. In standard statistical practice, ddof=1 provides an unbiased estimator of the variance of the infinite population. ddof=0 provides a maximum likelihood estimate of the variance for normally distributed variables. The standard deviation computed in this function is the square root of the estimated variance, so even with ddof=1, it will not be an unbiased estimate of the standard deviation per se.

Note that, for complex numbers, std takes the absolute value before squaring, so that the result is always real and nonnegative.

For floating-point input, the std is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-accuracy accumulator using the dtype keyword can alleviate this issue.

Examples

>>> a = np.array([[1, 2], [3, 4]])
>>> np.std(a)
1.1180339887498949
>>> np.std(a, axis=0)
array([ 1.,  1.])
>>> np.std(a, axis=1)
array([ 0.5,  0.5])

In single precision, std() can be inaccurate:

>>> a = np.zeros((2, 512*512), dtype=np.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.std(a)
0.45000005

Computing the standard deviation in float64 is more accurate:

>>> np.std(a, dtype=np.float64)
0.44999999925494177
dask.array.sum(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Sum of array elements over a given axis.

Parameters:

a : array_like

Elements to sum.

axis : None or int or tuple of ints, optional

Axis or axes along which a sum is performed. The default, axis=None, will sum all of the elements of the input array. If axis is negative it counts from the last to the first axis.

New in version 1.7.0.

If axis is a tuple of ints, a sum is performed on all of the axes specified in the tuple instead of a single axis or all the axes as before.

dtype : dtype, optional

The type of the returned array and of the accumulator in which the elements are summed. The dtype of a is used by default unless a has an integer dtype of less precision than the default platform integer. In that case, if a is signed then the platform integer is used while if a is unsigned then an unsigned integer of the same precision as the platform integer is used.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the sum method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

sum_along_axis : ndarray

An array with the same shape as a, with the specified axis removed. If a is a 0-d array, or if axis is None, a scalar is returned. If an output array is specified, a reference to out is returned.

See also

ndarray.sum
Equivalent method.
cumsum
Cumulative sum of array elements.
trapz
Integration of array values using the composite trapezoidal rule.

mean, average

Notes

Arithmetic is modular when using integer types, and no error is raised on overflow.

The sum of an empty array is the neutral element 0:

>>> np.sum([])
0.0

Examples

>>> np.sum([0.5, 1.5])
2.0
>>> np.sum([0.5, 0.7, 0.2, 1.5], dtype=np.int32)
1
>>> np.sum([[0, 1], [0, 5]])
6
>>> np.sum([[0, 1], [0, 5]], axis=0)
array([0, 6])
>>> np.sum([[0, 1], [0, 5]], axis=1)
array([1, 5])

If the accumulator is too small, overflow occurs:

>>> np.ones(128, dtype=np.int8).sum(dtype=np.int8)
-128
dask.array.take(a, indices, axis=None, out=None, mode='raise')

Take elements from an array along an axis.

This function does the same thing as “fancy” indexing (indexing arrays using arrays); however, it can be easier to use if you need elements along a given axis.

Parameters:

a : array_like

The source array.

indices : array_like

The indices of the values to extract.

New in version 1.8.0.

Also allow scalars for indices.

axis : int, optional

The axis over which to select values. By default, the flattened input array is used.

out : ndarray, optional

If provided, the result will be placed in this array. It should be of the appropriate shape and dtype.

mode : {‘raise’, ‘wrap’, ‘clip’}, optional

Specifies how out-of-bounds indices will behave.

  • ‘raise’ – raise an error (default)
  • ‘wrap’ – wrap around
  • ‘clip’ – clip to the range

‘clip’ mode means that all indices that are too large are replaced by the index that addresses the last element along that axis. Note that this disables indexing with negative numbers.
Returns:

subarray : ndarray

The returned array has the same type as a.

See also

compress
Take elements using a boolean mask
ndarray.take
equivalent method

Examples

>>> a = [4, 3, 5, 7, 6, 8]
>>> indices = [0, 1, 4]
>>> np.take(a, indices)
array([4, 3, 6])

In this example if a is an ndarray, “fancy” indexing can be used.

>>> a = np.array(a)
>>> a[indices]
array([4, 3, 6])

If indices is not one dimensional, the output also has these dimensions.

>>> np.take(a, [[0, 1], [2, 3]])
array([[4, 3],
       [5, 7]])
dask.array.tan(x[, out])

Compute tangent element-wise.

Equivalent to np.sin(x)/np.cos(x) element-wise.

Parameters:

x : array_like

Input array.

out : ndarray, optional

Output array of same shape as x.
Returns:

y : ndarray

The corresponding tangent values.
Raises:

ValueError: invalid return array shape

if out is provided and out.shape != x.shape (See Examples)

Notes

If out is provided, the function writes the result into it, and returns a reference to out. (See Examples)

References

M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions. New York, NY: Dover, 1972.

Examples

>>> from math import pi  
>>> np.tan(np.array([-pi,pi/2,pi]))  
array([  1.22460635e-16,   1.63317787e+16,  -1.22460635e-16])
>>>
>>> # Example of providing the optional output parameter illustrating
>>> # that what is returned is a reference to said parameter
>>> out2 = np.cos([0.1], out1)  
>>> out2 is out1  
True
>>>
>>> # Example of ValueError due to provision of shape mis-matched `out`
>>> np.cos(np.zeros((3,3)),np.zeros((2,2)))  
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: invalid return array shape
dask.array.tanh(x[, out])

Compute hyperbolic tangent element-wise.

Equivalent to np.sinh(x)/np.cosh(x) or -1j * np.tan(1j*x).

Parameters:

x : array_like

Input array.

out : ndarray, optional

Output array of same shape as x.
Returns:

y : ndarray

The corresponding hyperbolic tangent values.
Raises:

ValueError: invalid return array shape

if out is provided and out.shape != x.shape (See Examples)

Notes

If out is provided, the function writes the result into it, and returns a reference to out. (See Examples)

References

[R241242]M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions. New York, NY: Dover, 1972, pg. 83. http://www.math.sfu.ca/~cbm/aands/
[R242242]Wikipedia, “Hyperbolic function”, http://en.wikipedia.org/wiki/Hyperbolic_function

Examples

>>> np.tanh((0, np.pi*1j, np.pi*1j/2))  
array([ 0. +0.00000000e+00j,  0. -1.22460635e-16j,  0. +1.63317787e+16j])

>>> # Example of providing the optional output parameter illustrating
>>> # that what is returned is a reference to said parameter
>>> out2 = np.tanh([0.1], out1)  
>>> out2 is out1  
True

>>> # Example of ValueError due to provision of shape mis-matched `out`
>>> np.tanh(np.zeros((3,3)),np.zeros((2,2)))  
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: invalid return array shape
dask.array.tensordot(a, b, axes=2)

Compute tensor dot product along specified axes for arrays >= 1-D.

Given two tensors (arrays of dimension greater than or equal to one), a and b, and an array_like object containing two array_like objects, (a_axes, b_axes), sum the products of a‘s and b‘s elements (components) over the axes specified by a_axes and b_axes. The third argument can be a single non-negative integer_like scalar, N; if it is such, then the last N dimensions of a and the first N dimensions of b are summed over.

Parameters:

a, b : array_like, len(shape) >= 1

Tensors to “dot”.

axes : int or (2,) array_like

  • integer_like If an int N, sum over the last N axes of a and the first N axes of b in order. The sizes of the corresponding axes must match.
  • (2,) array_like Or, a list of axes to be summed over, first sequence applying to a, second to b. Both elements array_like must be of the same length.

See also

dot, einsum

Notes

Three common use cases are:
  • axes = 0 : tensor product \(a\otimes b\)
  • axes = 1 : tensor dot product \(a\cdot b\)
  • axes = 2 : (default) tensor double contraction \(a:b\)

When axes is integer_like, the sequence for evaluation will be: first the -Nth axis in a and 0th axis in b, and the -1th axis in a and Nth axis in b last.

When there is more than one axis to sum over - and they are not the last (first) axes of a (b) - the argument axes should consist of two sequences of the same length, with the first axis to sum over given first in both sequences, the second axis second, and so forth.

Examples

A “traditional” example:

>>> a = np.arange(60.).reshape(3,4,5)
>>> b = np.arange(24.).reshape(4,3,2)
>>> c = np.tensordot(a,b, axes=([1,0],[0,1]))
>>> c.shape
(5, 2)
>>> c
array([[ 4400.,  4730.],
       [ 4532.,  4874.],
       [ 4664.,  5018.],
       [ 4796.,  5162.],
       [ 4928.,  5306.]])
>>> # A slower but equivalent way of computing the same...
>>> d = np.zeros((5,2))
>>> for i in range(5):
...   for j in range(2):
...     for k in range(3):
...       for n in range(4):
...         d[i,j] += a[k,n,i] * b[n,k,j]
>>> c == d
array([[ True,  True],
       [ True,  True],
       [ True,  True],
       [ True,  True],
       [ True,  True]], dtype=bool)

An extended example taking advantage of the overloading of + and *:

>>> a = np.array(range(1, 9))
>>> a.shape = (2, 2, 2)
>>> A = np.array(('a', 'b', 'c', 'd'), dtype=object)
>>> A.shape = (2, 2)
>>> a; A
array([[[1, 2],
        [3, 4]],
       [[5, 6],
        [7, 8]]])
array([[a, b],
       [c, d]], dtype=object)

>>> np.tensordot(a, A) # third argument default is 2 for double-contraction
array([abbcccdddd, aaaaabbbbbbcccccccdddddddd], dtype=object)

>>> np.tensordot(a, A, 1)
array([[[acc, bdd],
        [aaacccc, bbbdddd]],
       [[aaaaacccccc, bbbbbdddddd],
        [aaaaaaacccccccc, bbbbbbbdddddddd]]], dtype=object)

>>> np.tensordot(a, A, 0) # tensor product (result too long to incl.)
array([[[[[a, b],
          [c, d]],
          ...

>>> np.tensordot(a, A, (0, 1))
array([[[abbbbb, cddddd],
        [aabbbbbb, ccdddddd]],
       [[aaabbbbbbb, cccddddddd],
        [aaaabbbbbbbb, ccccdddddddd]]], dtype=object)

>>> np.tensordot(a, A, (2, 1))
array([[[abb, cdd],
        [aaabbbb, cccdddd]],
       [[aaaaabbbbbb, cccccdddddd],
        [aaaaaaabbbbbbbb, cccccccdddddddd]]], dtype=object)

>>> np.tensordot(a, A, ((0, 1), (0, 1)))
array([abbbcccccddddddd, aabbbbccccccdddddddd], dtype=object)

>>> np.tensordot(a, A, ((2, 1), (1, 0)))
array([acccbbdddd, aaaaacccccccbbbbbbdddddddd], dtype=object)
dask.array.tile(A, reps)

Construct an array by repeating A the number of times given by reps.

If reps has length d, the result will have dimension of max(d, A.ndim).

If A.ndim < d, A is promoted to be d-dimensional by prepending new axes. So a shape (3,) array is promoted to (1, 3) for 2-D replication, or shape (1, 1, 3) for 3-D replication. If this is not the desired behavior, promote A to d-dimensions manually before calling this function.

If A.ndim > d, reps is promoted to A.ndim by pre-pending 1’s to it. Thus for an A of shape (2, 3, 4, 5), a reps of (2, 2) is treated as (1, 1, 2, 2).

Note : Although tile may be used for broadcasting, it is strongly recommended to use numpy’s broadcasting operations and functions.

Parameters:

A : array_like

The input array.

reps : array_like

The number of repetitions of A along each axis.
Returns:

c : ndarray

The tiled output array.

See also

repeat
Repeat elements of an array.
broadcast_to
Broadcast an array to a new shape

Examples

>>> a = np.array([0, 1, 2])
>>> np.tile(a, 2)
array([0, 1, 2, 0, 1, 2])
>>> np.tile(a, (2, 2))
array([[0, 1, 2, 0, 1, 2],
       [0, 1, 2, 0, 1, 2]])
>>> np.tile(a, (2, 1, 2))
array([[[0, 1, 2, 0, 1, 2]],
       [[0, 1, 2, 0, 1, 2]]])

>>> b = np.array([[1, 2], [3, 4]])
>>> np.tile(b, 2)
array([[1, 2, 1, 2],
       [3, 4, 3, 4]])
>>> np.tile(b, (2, 1))
array([[1, 2],
       [3, 4],
       [1, 2],
       [3, 4]])

>>> c = np.array([1,2,3,4])
>>> np.tile(c,(4,1))
array([[1, 2, 3, 4],
       [1, 2, 3, 4],
       [1, 2, 3, 4],
       [1, 2, 3, 4]])
dask.array.topk(k, x)

The top k elements of an array

Returns the k greatest elements of the array in sorted order. Only works on arrays of a single dimension.

This assumes that k is small. All results will be returned in a single chunk.

Examples

>>> x = np.array([5, 1, 3, 6])
>>> d = from_array(x, chunks=2)
>>> d.topk(2).compute()
array([6, 5])
dask.array.transpose(a, axes=None)

Permute the dimensions of an array.

Parameters:

a : array_like

Input array.

axes : list of ints, optional

By default, reverse the dimensions, otherwise permute the axes according to the values given.
Returns:

p : ndarray

a with its axes permuted. A view is returned whenever possible.

See also

moveaxis, argsort

Notes

Use transpose(a, argsort(axes)) to invert the transposition of tensors when using the axes keyword argument.

Transposing a 1-D array returns an unchanged view of the original array.

Examples

>>> x = np.arange(4).reshape((2,2))
>>> x
array([[0, 1],
       [2, 3]])

>>> np.transpose(x)
array([[0, 2],
       [1, 3]])

>>> x = np.ones((1, 2, 3))
>>> np.transpose(x, (1, 0, 2)).shape
(2, 1, 3)
dask.array.tril(m, k=0)

Lower triangle of an array with elements above the k-th diagonal zeroed.

Parameters:

m : array_like, shape (M, M)

Input array.

k : int, optional

Diagonal above which to zero elements. k = 0 (the default) is the main diagonal, k < 0 is below it and k > 0 is above.
Returns:

tril : ndarray, shape (M, M)

Lower triangle of m, of same shape and data-type as m.

See also

triu
upper triangle of an array
dask.array.triu(m, k=0)

Upper triangle of an array with elements above the k-th diagonal zeroed.

Parameters:

m : array_like, shape (M, N)

Input array.

k : int, optional

Diagonal above which to zero elements. k = 0 (the default) is the main diagonal, k < 0 is below it and k > 0 is above.
Returns:

triu : ndarray, shape (M, N)

Upper triangle of m, of same shape and data-type as m.

See also

tril
lower triangle of an array
dask.array.trunc(x[, out])

Return the truncated value of the input, element-wise.

The truncated value of the scalar x is the nearest integer i which is closer to zero than x is. In short, the fractional part of the signed number x is discarded.

Parameters:

x : array_like

Input data.
Returns:

y : ndarray or scalar

The truncated value of each element in x.

See also

ceil, floor, rint

Notes

New in version 1.3.0.

Examples

>>> a = np.array([-1.7, -1.5, -0.2, 0.2, 1.5, 1.7, 2.0])  
>>> np.trunc(a)  
array([-1., -1., -0.,  0.,  1.,  1.,  2.])
dask.array.unique(ar, return_index=False, return_inverse=False, return_counts=False)

Find the unique elements of an array.

Returns the sorted unique elements of an array. There are three optional outputs in addition to the unique elements: the indices of the input array that give the unique values, the indices of the unique array that reconstruct the input array, and the number of times each unique value comes up in the input array.

Parameters:

ar : array_like

Input array. This will be flattened if it is not already 1-D.

return_index : bool, optional

If True, also return the indices of ar that result in the unique array.

return_inverse : bool, optional

If True, also return the indices of the unique array that can be used to reconstruct ar.

return_counts : bool, optional

If True, also return the number of times each unique value comes up in ar.

New in version 1.9.0.
Returns:

unique : ndarray

The sorted unique values.

unique_indices : ndarray, optional

The indices of the first occurrences of the unique values in the (flattened) original array. Only provided if return_index is True.

unique_inverse : ndarray, optional

The indices to reconstruct the (flattened) original array from the unique array. Only provided if return_inverse is True.

unique_counts : ndarray, optional

The number of times each of the unique values comes up in the original array. Only provided if return_counts is True.

New in version 1.9.0.

See also

numpy.lib.arraysetops
Module with a number of other functions for performing set operations on arrays.

Examples

>>> np.unique([1, 1, 2, 2, 3, 3])
array([1, 2, 3])
>>> a = np.array([[1, 1], [2, 3]])
>>> np.unique(a)
array([1, 2, 3])

Return the indices of the original array that give the unique values:

>>> a = np.array(['a', 'b', 'b', 'c', 'a'])
>>> u, indices = np.unique(a, return_index=True)
>>> u
array(['a', 'b', 'c'],
       dtype='|S1')
>>> indices
array([0, 1, 3])
>>> a[indices]
array(['a', 'b', 'c'],
       dtype='|S1')

Reconstruct the input array from the unique values:

>>> a = np.array([1, 2, 6, 4, 2, 3, 2])
>>> u, indices = np.unique(a, return_inverse=True)
>>> u
array([1, 2, 3, 4, 6])
>>> indices
array([0, 1, 4, 3, 1, 2, 1])
>>> u[indices]
array([1, 2, 6, 4, 2, 3, 2])
dask.array.var(a, axis=None, dtype=None, out=None, ddof=0, keepdims=<class 'numpy._globals._NoValue'>)

Compute the variance along the specified axis.

Returns the variance of the array elements, a measure of the spread of a distribution. The variance is computed for the flattened array by default, otherwise over the specified axis.

Parameters:

a : array_like

Array containing numbers whose variance is desired. If a is not an array, a conversion is attempted.

axis : None or int or tuple of ints, optional

Axis or axes along which the variance is computed. The default is to compute the variance of the flattened array.

New in version 1.7.0.

If this is a tuple of ints, a variance is performed over multiple axes, instead of a single axis or all the axes as before.

dtype : data-type, optional

Type to use in computing the variance. For arrays of integer type the default is float32; for arrays of float types it is the same as the array type.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output, but the type is cast if necessary.

ddof : int, optional

“Delta Degrees of Freedom”: the divisor used in the calculation is N - ddof, where N represents the number of elements. By default ddof is zero.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the var method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

variance : ndarray, see dtype parameter above

If out=None, returns a new array containing the variance; otherwise, a reference to the output array is returned.

See also

std, mean, nanmean, nanstd, nanvar

numpy.doc.ufuncs
Section “Output arguments”

Notes

The variance is the average of the squared deviations from the mean, i.e., var = mean(abs(x - x.mean())**2).

The mean is normally calculated as x.sum() / N, where N = len(x). If, however, ddof is specified, the divisor N - ddof is used instead. In standard statistical practice, ddof=1 provides an unbiased estimator of the variance of a hypothetical infinite population. ddof=0 provides a maximum likelihood estimate of the variance for normally distributed variables.

Note that for complex numbers, the absolute value is taken before squaring, so that the result is always real and nonnegative.

For floating-point input, the variance is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-accuracy accumulator using the dtype keyword can alleviate this issue.

Examples

>>> a = np.array([[1, 2], [3, 4]])
>>> np.var(a)
1.25
>>> np.var(a, axis=0)
array([ 1.,  1.])
>>> np.var(a, axis=1)
array([ 0.25,  0.25])

In single precision, var() can be inaccurate:

>>> a = np.zeros((2, 512*512), dtype=np.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.var(a)
0.20250003

Computing the variance in float64 is more accurate:

>>> np.var(a, dtype=np.float64)
0.20249999932944759
>>> ((1-0.55)**2 + (0.1-0.55)**2)/2
0.2025
dask.array.vnorm(a, ord=None, axis=None, dtype=None, keepdims=False, split_every=None, out=None)

Vector norm

See np.linalg.norm

dask.array.vstack(tup)

Stack arrays in sequence vertically (row wise).

Take a sequence of arrays and stack them vertically to make a single array. Rebuild arrays divided by vsplit.

This function continues to be supported for backward compatibility, but you should prefer np.concatenate or np.stack. The np.stack function was added in NumPy 1.10.

Parameters:

tup : sequence of ndarrays

Tuple containing arrays to be stacked. The arrays must have the same shape along all but the first axis.
Returns:

stacked : ndarray

The array formed by stacking the given arrays.

See also

stack
Join a sequence of arrays along a new axis.
hstack
Stack arrays in sequence horizontally (column wise).
dstack
Stack arrays in sequence depth wise (along third dimension).
concatenate
Join a sequence of arrays along an existing axis.
vsplit
Split array into a list of multiple sub-arrays vertically.

Notes

Equivalent to np.concatenate(tup, axis=0) if tup contains arrays that are at least 2-dimensional.

Examples

>>> a = np.array([1, 2, 3])
>>> b = np.array([2, 3, 4])
>>> np.vstack((a,b))
array([[1, 2, 3],
       [2, 3, 4]])

>>> a = np.array([[1], [2], [3]])
>>> b = np.array([[2], [3], [4]])
>>> np.vstack((a,b))
array([[1],
       [2],
       [3],
       [2],
       [3],
       [4]])
dask.array.where(condition[, x, y])

Return elements, either from x or y, depending on condition.

If only condition is given, return condition.nonzero().

Parameters:

condition : array_like, bool

When True, yield x, otherwise yield y.

x, y : array_like, optional

Values from which to choose. x and y need to have the same shape as condition.
Returns:

out : ndarray or tuple of ndarrays

If both x and y are specified, the output array contains elements of x where condition is True, and elements from y elsewhere.

If only condition is given, return the tuple condition.nonzero(), the indices where condition is True.

See also

nonzero, choose

Notes

If x and y are given and input arrays are 1-D, where is equivalent to:

[xv if c else yv for (c,xv,yv) in zip(condition,x,y)]

Examples

>>> np.where([[True, False], [True, True]],
...          [[1, 2], [3, 4]],
...          [[9, 8], [7, 6]])
array([[1, 8],
       [3, 4]])

>>> np.where([[0, 1], [1, 0]])
(array([0, 1]), array([1, 0]))

>>> x = np.arange(9.).reshape(3, 3)
>>> np.where( x > 5 )
(array([2, 2, 2]), array([0, 1, 2]))
>>> x[np.where( x > 3.0 )]               # Note: result is 1D.
array([ 4.,  5.,  6.,  7.,  8.])
>>> np.where(x < 5, x, -1)               # Note: broadcasting.
array([[ 0.,  1.,  2.],
       [ 3.,  4., -1.],
       [-1., -1., -1.]])

Find the indices of elements of x that are in goodvalues.

>>> goodvalues = [3, 4, 7]
>>> ix = np.in1d(x.ravel(), goodvalues).reshape(x.shape)
>>> ix
array([[False, False, False],
       [ True,  True, False],
       [False,  True, False]], dtype=bool)
>>> np.where(ix)
(array([1, 1, 2]), array([0, 1, 1]))
dask.array.zeros()

Blocked variant of zeros

Follows the signature of zeros exactly except that it also requires a keyword argument chunks=(...)

Original signature follows below. zeros(shape, dtype=float, order=’C’)

Return a new array of given shape and type, filled with zeros.

Parameters:

shape : int or sequence of ints

Shape of the new array, e.g., (2, 3) or 2.

dtype : data-type, optional

The desired data-type for the array, e.g., numpy.int8. Default is numpy.float64.

order : {‘C’, ‘F’}, optional

Whether to store multidimensional data in C- or Fortran-contiguous (row- or column-wise) order in memory.
Returns:

out : ndarray

Array of zeros with the given shape, dtype, and order.

See also

zeros_like
Return an array of zeros with shape and type of input.
ones_like
Return an array of ones with shape and type of input.
empty_like
Return an empty array with shape and type of input.
ones
Return a new array setting values to one.
empty
Return a new uninitialized array.

Examples

>>> np.zeros(5)
array([ 0.,  0.,  0.,  0.,  0.])

>>> np.zeros((5,), dtype=np.int)
array([0, 0, 0, 0, 0])

>>> np.zeros((2, 1))
array([[ 0.],
       [ 0.]])

>>> s = (2,2)
>>> np.zeros(s)
array([[ 0.,  0.],
       [ 0.,  0.]])

>>> np.zeros((2,), dtype=[('x', 'i4'), ('y', 'i4')]) # custom dtype
array([(0, 0), (0, 0)],
      dtype=[('x', '<i4'), ('y', '<i4')])
dask.array.linalg.cholesky(a, lower=False)

Returns the Cholesky decomposition, \(A = L L^*\) or \(A = U^* U\) of a Hermitian positive-definite matrix A.

Parameters:

a : (M, M) array_like

Matrix to be decomposed

lower : bool, optional

Whether to compute the upper or lower triangular Cholesky factorization. Default is upper-triangular.
Returns:

c : (M, M) Array

Upper- or lower-triangular Cholesky factor of a.
dask.array.linalg.inv(a)

Compute the inverse of a matrix with LU decomposition and forward / backward substitutions.

Parameters:

a : array_like

Square matrix to be inverted.
Returns:

ainv : Array

Inverse of the matrix a.
dask.array.linalg.lstsq(a, b)

Return the least-squares solution to a linear matrix equation using QR decomposition.

Solves the equation a x = b by computing a vector x that minimizes the Euclidean 2-norm || b - a x ||^2. The equation may be under-, well-, or over- determined (i.e., the number of linearly independent rows of a can be less than, equal to, or greater than its number of linearly independent columns). If a is square and of full rank, then x (but for round-off error) is the “exact” solution of the equation.

Parameters:

a : (M, N) array_like

“Coefficient” matrix.

b : (M,) array_like

Ordinate or “dependent variable” values.
Returns:

x : (N,) Array

Least-squares solution. If b is two-dimensional, the solutions are in the K columns of x.

residuals : (1,) Array

Sums of residuals; squared Euclidean 2-norm for each column in b - a*x.

rank : Array

Rank of matrix a.

s : (min(M, N),) Array

Singular values of a.
dask.array.linalg.lu(a)

Compute the lu decomposition of a matrix.

Returns:

p: Array, permutation matrix

l: Array, lower triangular matrix with unit diagonal.

u: Array, upper triangular matrix

Examples

>>> p, l, u = da.linalg.lu(x)
dask.array.linalg.qr(a, name=None)

Compute the qr factorization of a matrix.

Returns:

q: Array, orthonormal

r: Array, upper-triangular

See also

np.linalg.qr
Equivalent NumPy Operation
dask.array.linalg.tsqr
Actual implementation with citation

Examples

>>> q, r = da.linalg.qr(x)
dask.array.linalg.solve(a, b, sym_pos=False)

Solve the equation a x = b for x. By default, use LU decomposition and forward / backward substitutions. When sym_pos is True, use Cholesky decomposition.

Parameters:

a : (M, M) array_like

A square matrix.

b : (M,) or (M, N) array_like

Right-hand side matrix in a x = b.

sym_pos : bool

Assume a is symmetric and positive definite. If True, use Cholesky decomposition.
Returns:

x : (M,) or (M, N) Array

Solution to the system a x = b. Shape of the return matches the shape of b.
dask.array.linalg.solve_triangular(a, b, lower=False)

Solve the equation a x = b for x, assuming a is a triangular matrix.

Parameters:

a : (M, M) array_like

A triangular matrix

b : (M,) or (M, N) array_like

Right-hand side matrix in a x = b

lower : bool, optional

Use only data contained in the lower triangle of a. Default is to use upper triangle.
Returns:

x : (M,) or (M, N) array

Solution to the system a x = b. Shape of return matches b.
dask.array.linalg.svd(a, name=None)

Compute the singular value decomposition of a matrix.

Returns:

u: Array, unitary / orthogonal

s: Array, singular values in decreasing order (largest first)

v: Array, unitary / orthogonal

See also

np.linalg.svd
Equivalent NumPy Operation
dask.array.linalg.tsqr
Actual implementation with citation

Examples

>>> u, s, v = da.linalg.svd(x)
dask.array.linalg.svd_compressed(a, k, n_power_iter=0, seed=None, name=None)

Randomly compressed rank-k thin Singular Value Decomposition.

This computes the approximate singular value decomposition of a large array. This algorithm is generally faster than the normal algorithm but does not provide exact results. One can balance between performance and accuracy with input parameters (see below).

Parameters:

a: Array

Input array

k: int

Rank of the desired thin SVD decomposition.

n_power_iter: int

Number of power iterations, useful when the singular values decay slowly. Error decreases exponentially as n_power_iter increases. In practice, set n_power_iter <= 4.
Returns:

u: Array, unitary / orthogonal

s: Array, singular values in decreasing order (largest first)

v: Array, unitary / orthogonal

References

N. Halko, P. G. Martinsson, and J. A. Tropp. Finding structure with randomness: Probabilistic algorithms for constructing approximate matrix decompositions. SIAM Rev., Survey and Review section, Vol. 53, num. 2, pp. 217-288, June 2011 http://arxiv.org/abs/0909.4061

Examples

>>> u, s, vt = svd_compressed(x, 20)
dask.array.linalg.tsqr(data, name=None, compute_svd=False)

Direct Tall-and-Skinny QR algorithm

As presented in:

A. Benson, D. Gleich, and J. Demmel. Direct QR factorizations for tall-and-skinny matrices in MapReduce architectures. IEEE International Conference on Big Data, 2013. http://arxiv.org/abs/1301.1071

This algorithm is used to compute both the QR decomposition and the Singular Value Decomposition. It requires that the input array have a single column of blocks, each of which fit in memory.

If blocks are of size (n, k) then this algorithm has memory use that scales as n**2 * k * nthreads.

Parameters:

data: Array

compute_svd: bool

Whether to compute the SVD rather than the QR decomposition

See also

dask.array.linalg.qr, dask.array.linalg.svd

dask.array.ghost.ghost(x, depth, boundary)

Share boundaries between neighboring blocks

Parameters:

x: da.Array

A dask array

depth: dict

The size of the shared boundary per axis

boundary: dict

The boundary condition on each axis. Options are ‘reflect’, ‘periodic’, ‘nearest’, ‘none’, or an array value. Such a value will fill the boundary with that value.

The depth input informs how many cells to overlap between neighboring

blocks ``{0: 2, 2: 5}`` means share two cells in 0 axis, 5 cells in 2 axis.

Axes missing from this input will not be overlapped.

Examples

>>> import numpy as np
>>> import dask.array as da

>>> x = np.arange(64).reshape((8, 8))
>>> d = da.from_array(x, chunks=(4, 4))
>>> d.chunks
((4, 4), (4, 4))

>>> g = da.ghost.ghost(d, depth={0: 2, 1: 1},
...                       boundary={0: 100, 1: 'reflect'})
>>> g.chunks
((8, 8), (6, 6))

>>> np.array(g)
array([[100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
       [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
       [  0,   0,   1,   2,   3,   4,   3,   4,   5,   6,   7,   7],
       [  8,   8,   9,  10,  11,  12,  11,  12,  13,  14,  15,  15],
       [ 16,  16,  17,  18,  19,  20,  19,  20,  21,  22,  23,  23],
       [ 24,  24,  25,  26,  27,  28,  27,  28,  29,  30,  31,  31],
       [ 32,  32,  33,  34,  35,  36,  35,  36,  37,  38,  39,  39],
       [ 40,  40,  41,  42,  43,  44,  43,  44,  45,  46,  47,  47],
       [ 16,  16,  17,  18,  19,  20,  19,  20,  21,  22,  23,  23],
       [ 24,  24,  25,  26,  27,  28,  27,  28,  29,  30,  31,  31],
       [ 32,  32,  33,  34,  35,  36,  35,  36,  37,  38,  39,  39],
       [ 40,  40,  41,  42,  43,  44,  43,  44,  45,  46,  47,  47],
       [ 48,  48,  49,  50,  51,  52,  51,  52,  53,  54,  55,  55],
       [ 56,  56,  57,  58,  59,  60,  59,  60,  61,  62,  63,  63],
       [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
       [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100]])
dask.array.ghost.map_overlap(x, func, depth, boundary=None, trim=True, **kwargs)
dask.array.from_array(x, chunks, name=None, lock=False, asarray=True, fancy=True, getitem=None)

Create dask array from something that looks like an array

Input must have a .shape and support numpy-style slicing.

Parameters:

x : array_like

chunks : int, tuple

How to chunk the array. Must be one of the following forms: - A blocksize like 1000. - A blockshape like (1000, 1000). - Explicit sizes of all blocks along all dimensions

like ((1000, 1000, 500), (400, 400)).

name : str, optional

The key name to use for the array. Defaults to a hash of x. Use name=False to generate a random name instead of hashing (fast)

lock : bool or Lock, optional

If x doesn’t support concurrent reads then provide a lock here, or pass in True to have dask.array create one for you.

asarray : bool, optional

If True (default), then chunks will be converted to instances of ndarray. Set to False to pass passed chunks through unchanged.

fancy : bool, optional

If x doesn’t support fancy indexing (e.g. indexing with lists or arrays) then set to False. Default is True.

Examples

>>> x = h5py.File('...')['/data/path']  
>>> a = da.from_array(x, chunks=(1000, 1000))

If your underlying datastore does not support concurrent reads then include the lock=True keyword argument or lock=mylock if you want multiple arrays to coordinate around the same lock.

>>> a = da.from_array(x, chunks=(1000, 1000), lock=True)
dask.array.from_delayed(value, shape, dtype, name=None)

Create a dask array from a dask delayed value

This routine is useful for constructing dask arrays in an ad-hoc fashion using dask delayed, particularly when combined with stack and concatenate.

The dask array will consist of a single chunk.

Examples

>>> from dask import delayed
>>> value = delayed(np.ones)(5)
>>> array = from_delayed(value, (5,), float)
>>> array
dask.array<from-value, shape=(5,), dtype=float64, chunksize=(5,)>
>>> array.compute()
array([ 1.,  1.,  1.,  1.,  1.])
dask.array.from_npy_stack(dirname, mmap_mode='r')

Load dask array from stack of npy files

See da.to_npy_stack for docstring

Parameters:

dirname: string

Directory of .npy files

mmap_mode: (None or ‘r’)

Read data in memory map mode
dask.array.store(sources, targets, lock=True, regions=None, compute=True, **kwargs)

Store dask arrays in array-like objects, overwrite data in target

This stores dask arrays into object that supports numpy-style setitem indexing. It stores values chunk by chunk so that it does not have to fill up memory. For best performance you can align the block size of the storage target with the block size of your array.

If your data fits in memory then you may prefer calling np.array(myarray) instead.

Parameters:

sources: Array or iterable of Arrays

targets: array-like or iterable of array-likes

These should support setitem syntax target[10:20] = ...

lock: boolean or threading.Lock, optional

Whether or not to lock the data stores while storing. Pass True (lock each file individually), False (don’t lock) or a particular threading.Lock object to be shared among all writes.

regions: tuple of slices or iterable of tuple of slices

Each region tuple in regions should be such that target[region].shape = source.shape for the corresponding source and target in sources and targets, respectively.

compute: boolean, optional

If true compute immediately, return dask.delayed.Delayed otherwise

Examples

>>> x = ...  

>>> import h5py  
>>> f = h5py.File('myfile.hdf5')  
>>> dset = f.create_dataset('/data', shape=x.shape,
...                                  chunks=x.chunks,
...                                  dtype='f8')  

>>> store(x, dset)

Alternatively store many arrays at the same time

>>> store([x, y, z], [dset1, dset2, dset3])
dask.array.to_hdf5(filename, *args, **kwargs)

Store arrays in HDF5 file

This saves several dask arrays into several datapaths in an HDF5 file. It creates the necessary datasets and handles clean file opening/closing.

>>> da.to_hdf5('myfile.hdf5', '/x', x)

or

>>> da.to_hdf5('myfile.hdf5', {'/x': x, '/y': y})

Optionally provide arguments as though to h5py.File.create_dataset

>>> da.to_hdf5('myfile.hdf5', '/x', x, compression='lzf', shuffle=True)

This can also be used as a method on a single Array

>>> x.to_hdf5('myfile.hdf5', '/x')

See also

da.store, h5py.File.create_dataset

dask.array.to_npy_stack(dirname, x, axis=0)

Write dask array to a stack of .npy files

This partitions the dask.array along one axis and stores each block along that axis as a single .npy file in the specified directory

See also

from_npy_stack

Examples

>>> x = da.ones((5, 10, 10), chunks=(2, 4, 4))  
>>> da.to_npy_stack('data/', x, axis=0)
$ tree data/ data/ |– 0.npy |– 1.npy |– 2.npy |– info

The .npy files store numpy arrays for x[0:2], x[2:4], and x[4:5] respectively, as is specified by the chunk size along the zeroth axis. The info file stores the dtype, chunks, and axis information of the array.

You can load these stacks with the da.from_npy_stack function.

>>> y = da.from_npy_stack('data/')
dask.array.fft.fft_wrap(fft_func, kind=None, dtype=None)

Wrap 1D complex FFT functions

Takes a function that behaves like numpy.fft functions and a specified kind to match it to that are named after the functions in the numpy.fft API.

Supported kinds include:

  • fft
  • ifft
  • rfft
  • irfft
  • hfft
  • ihfft

Examples

>>> parallel_fft = fft_wrap(np.fft.fft)
>>> parallel_ifft = fft_wrap(np.fft.ifft)
dask.array.fft.fft(a, n=None, axis=None)

Wrapping of numpy.fft.fftpack.fft

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.fft docstring follows below:

Compute the one-dimensional discrete Fourier Transform.

This function computes the one-dimensional n-point discrete Fourier Transform (DFT) with the efficient Fast Fourier Transform (FFT) algorithm [CT].

Parameters:

a : array_like

Input array, can be complex.

n : int, optional

Length of the transformed axis of the output. If n is smaller than the length of the input, the input is cropped. If it is larger, the input is padded with zeros. If n is not given, the length of the input along the axis specified by axis is used.

axis : int, optional

Axis over which to compute the FFT. If not given, the last axis is used.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axis indicated by axis, or the last one if axis is not specified.
Raises:

IndexError

if axes is larger than the last axis of a.

See also

numpy.fft
for definition of the DFT and conventions used.
ifft
The inverse of fft.
fft2
The two-dimensional FFT.
fftn
The n-dimensional FFT.
rfftn
The n-dimensional FFT of real input.
fftfreq
Frequency bins for given FFT parameters.

Notes

FFT (Fast Fourier Transform) refers to a way the discrete Fourier Transform (DFT) can be calculated efficiently, by using symmetries in the calculated terms. The symmetry is highest when n is a power of 2, and the transform is therefore most efficient for these sizes.

The DFT is defined, with the conventions used in this implementation, in the documentation for the numpy.fft module.

References

[CT244245]Cooley, James W., and John W. Tukey, 1965, “An algorithm for the machine calculation of complex Fourier series,” Math. Comput. 19: 297-301.

Examples

>>> np.fft.fft(np.exp(2j * np.pi * np.arange(8) / 8))
array([ -3.44505240e-16 +1.14383329e-17j,
         8.00000000e+00 -5.71092652e-15j,
         2.33482938e-16 +1.22460635e-16j,
         1.64863782e-15 +1.77635684e-15j,
         9.95839695e-17 +2.33482938e-16j,
         0.00000000e+00 +1.66837030e-15j,
         1.14383329e-17 +1.22460635e-16j,
         -1.64863782e-15 +1.77635684e-15j])

In this example, real input has an FFT which is Hermitian, i.e., symmetric in the real part and anti-symmetric in the imaginary part, as described in the numpy.fft documentation:

>>> import matplotlib.pyplot as plt
>>> t = np.arange(256)
>>> sp = np.fft.fft(np.sin(t))
>>> freq = np.fft.fftfreq(t.shape[-1])
>>> plt.plot(freq, sp.real, freq, sp.imag)
[<matplotlib.lines.Line2D object at 0x...>, <matplotlib.lines.Line2D object at 0x...>]
>>> plt.show()
dask.array.fft.fft2(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.fft2

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.fft2 docstring follows below:

Compute the 2-dimensional discrete Fourier Transform

This function computes the n-dimensional discrete Fourier Transform over any axes in an M-dimensional array by means of the Fast Fourier Transform (FFT). By default, the transform is computed over the last two axes of the input array, i.e., a 2-dimensional FFT.

Parameters:

a : array_like

Input array, can be complex

s : sequence of ints, optional

Shape (length of each transformed axis) of the output (s[0] refers to axis 0, s[1] to axis 1, etc.). This corresponds to n for fft(x, n). Along each axis, if the given shape is smaller than that of the input, the input is cropped. If it is larger, the input is padded with zeros. if s is not given, the shape of the input along the axes specified by axes is used.

axes : sequence of ints, optional

Axes over which to compute the FFT. If not given, the last two axes are used. A repeated index in axes means the transform over that axis is performed multiple times. A one-element sequence means that a one-dimensional FFT is performed.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axes indicated by axes, or the last two axes if axes is not given.
Raises:

ValueError

If s and axes have different length, or axes not given and len(s) != 2.

IndexError

If an element of axes is larger than than the number of axes of a.

See also

numpy.fft
Overall view of discrete Fourier transforms, with definitions and conventions used.
ifft2
The inverse two-dimensional FFT.
fft
The one-dimensional FFT.
fftn
The n-dimensional FFT.
fftshift
Shifts zero-frequency terms to the center of the array. For two-dimensional input, swaps first and third quadrants, and second and fourth quadrants.

Notes

fft2 is just fftn with a different default for axes.

The output, analogously to fft, contains the term for zero frequency in the low-order corner of the transformed axes, the positive frequency terms in the first half of these axes, the term for the Nyquist frequency in the middle of the axes and the negative frequency terms in the second half of the axes, in order of decreasingly negative frequency.

See fftn for details and a plotting example, and numpy.fft for definitions and conventions used.

Examples

>>> a = np.mgrid[:5, :5][0]
>>> np.fft.fft2(a)
array([[ 50.0 +0.j        ,   0.0 +0.j        ,   0.0 +0.j        ,
          0.0 +0.j        ,   0.0 +0.j        ],
       [-12.5+17.20477401j,   0.0 +0.j        ,   0.0 +0.j        ,
          0.0 +0.j        ,   0.0 +0.j        ],
       [-12.5 +4.0614962j ,   0.0 +0.j        ,   0.0 +0.j        ,
          0.0 +0.j        ,   0.0 +0.j        ],
       [-12.5 -4.0614962j ,   0.0 +0.j        ,   0.0 +0.j        ,
            0.0 +0.j        ,   0.0 +0.j        ],
       [-12.5-17.20477401j,   0.0 +0.j        ,   0.0 +0.j        ,
          0.0 +0.j        ,   0.0 +0.j        ]])
dask.array.fft.fftn(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.fftn

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.fftn docstring follows below:

Compute the N-dimensional discrete Fourier Transform.

This function computes the N-dimensional discrete Fourier Transform over any number of axes in an M-dimensional array by means of the Fast Fourier Transform (FFT).

Parameters:

a : array_like

Input array, can be complex.

s : sequence of ints, optional

Shape (length of each transformed axis) of the output (s[0] refers to axis 0, s[1] to axis 1, etc.). This corresponds to n for fft(x, n). Along any axis, if the given shape is smaller than that of the input, the input is cropped. If it is larger, the input is padded with zeros. if s is not given, the shape of the input along the axes specified by axes is used.

axes : sequence of ints, optional

Axes over which to compute the FFT. If not given, the last len(s) axes are used, or all axes if s is also not specified. Repeated indices in axes means that the transform over that axis is performed multiple times.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axes indicated by axes, or by a combination of s and a, as explained in the parameters section above.
Raises:

ValueError

If s and axes have different length.

IndexError

If an element of axes is larger than than the number of axes of a.

See also

numpy.fft
Overall view of discrete Fourier transforms, with definitions and conventions used.
ifftn
The inverse of fftn, the inverse n-dimensional FFT.
fft
The one-dimensional FFT, with definitions and conventions used.
rfftn
The n-dimensional FFT of real input.
fft2
The two-dimensional FFT.
fftshift
Shifts zero-frequency terms to centre of array

Notes

The output, analogously to fft, contains the term for zero frequency in the low-order corner of all axes, the positive frequency terms in the first half of all axes, the term for the Nyquist frequency in the middle of all axes and the negative frequency terms in the second half of all axes, in order of decreasingly negative frequency.

See numpy.fft for details, definitions and conventions used.

Examples

>>> a = np.mgrid[:3, :3, :3][0]
>>> np.fft.fftn(a, axes=(1, 2))
array([[[  0.+0.j,   0.+0.j,   0.+0.j],
        [  0.+0.j,   0.+0.j,   0.+0.j],
        [  0.+0.j,   0.+0.j,   0.+0.j]],
       [[  9.+0.j,   0.+0.j,   0.+0.j],
        [  0.+0.j,   0.+0.j,   0.+0.j],
        [  0.+0.j,   0.+0.j,   0.+0.j]],
       [[ 18.+0.j,   0.+0.j,   0.+0.j],
        [  0.+0.j,   0.+0.j,   0.+0.j],
        [  0.+0.j,   0.+0.j,   0.+0.j]]])
>>> np.fft.fftn(a, (2, 2), axes=(0, 1))
array([[[ 2.+0.j,  2.+0.j,  2.+0.j],
        [ 0.+0.j,  0.+0.j,  0.+0.j]],
       [[-2.+0.j, -2.+0.j, -2.+0.j],
        [ 0.+0.j,  0.+0.j,  0.+0.j]]])

>>> import matplotlib.pyplot as plt
>>> [X, Y] = np.meshgrid(2 * np.pi * np.arange(200) / 12,
...                      2 * np.pi * np.arange(200) / 34)
>>> S = np.sin(X) + np.cos(Y) + np.random.uniform(0, 1, X.shape)
>>> FS = np.fft.fftn(S)
>>> plt.imshow(np.log(np.abs(np.fft.fftshift(FS))**2))
<matplotlib.image.AxesImage object at 0x...>
>>> plt.show()
dask.array.fft.ifft(a, n=None, axis=None)

Wrapping of numpy.fft.fftpack.ifft

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.ifft docstring follows below:

Compute the one-dimensional inverse discrete Fourier Transform.

This function computes the inverse of the one-dimensional n-point discrete Fourier transform computed by fft. In other words, ifft(fft(a)) == a to within numerical accuracy. For a general description of the algorithm and definitions, see numpy.fft.

The input should be ordered in the same way as is returned by fft, i.e.,

  • a[0] should contain the zero frequency term,
  • a[1:n//2] should contain the positive-frequency terms,
  • a[n//2 + 1:] should contain the negative-frequency terms, in increasing order starting from the most negative frequency.

For an even number of input points, A[n//2] represents the sum of the values at the positive and negative Nyquist frequencies, as the two are aliased together. See numpy.fft for details.

Parameters:

a : array_like

Input array, can be complex.

n : int, optional

Length of the transformed axis of the output. If n is smaller than the length of the input, the input is cropped. If it is larger, the input is padded with zeros. If n is not given, the length of the input along the axis specified by axis is used. See notes about padding issues.

axis : int, optional

Axis over which to compute the inverse DFT. If not given, the last axis is used.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axis indicated by axis, or the last one if axis is not specified.
Raises:

IndexError

If axes is larger than the last axis of a.

See also

numpy.fft
An introduction, with definitions and general explanations.
fft
The one-dimensional (forward) FFT, of which ifft is the inverse
ifft2
The two-dimensional inverse FFT.
ifftn
The n-dimensional inverse FFT.

Notes

If the input parameter n is larger than the size of the input, the input is padded by appending zeros at the end. Even though this is the common approach, it might lead to surprising results. If a different padding is desired, it must be performed before calling ifft.

Examples

>>> np.fft.ifft([0, 4, 0, 0])
array([ 1.+0.j,  0.+1.j, -1.+0.j,  0.-1.j])

Create and plot a band-limited signal with random phases:

>>> import matplotlib.pyplot as plt
>>> t = np.arange(400)
>>> n = np.zeros((400,), dtype=complex)
>>> n[40:60] = np.exp(1j*np.random.uniform(0, 2*np.pi, (20,)))
>>> s = np.fft.ifft(n)
>>> plt.plot(t, s.real, 'b-', t, s.imag, 'r--')
...
>>> plt.legend(('real', 'imaginary'))
...
>>> plt.show()
dask.array.fft.ifft2(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.ifft2

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.ifft2 docstring follows below:

Compute the 2-dimensional inverse discrete Fourier Transform.

This function computes the inverse of the 2-dimensional discrete Fourier Transform over any number of axes in an M-dimensional array by means of the Fast Fourier Transform (FFT). In other words, ifft2(fft2(a)) == a to within numerical accuracy. By default, the inverse transform is computed over the last two axes of the input array.

The input, analogously to ifft, should be ordered in the same way as is returned by fft2, i.e. it should have the term for zero frequency in the low-order corner of the two axes, the positive frequency terms in the first half of these axes, the term for the Nyquist frequency in the middle of the axes and the negative frequency terms in the second half of both axes, in order of decreasingly negative frequency.

Parameters:

a : array_like

Input array, can be complex.

s : sequence of ints, optional

Shape (length of each axis) of the output (s[0] refers to axis 0, s[1] to axis 1, etc.). This corresponds to n for ifft(x, n). Along each axis, if the given shape is smaller than that of the input, the input is cropped. If it is larger, the input is padded with zeros. if s is not given, the shape of the input along the axes specified by axes is used. See notes for issue on ifft zero padding.

axes : sequence of ints, optional

Axes over which to compute the FFT. If not given, the last two axes are used. A repeated index in axes means the transform over that axis is performed multiple times. A one-element sequence means that a one-dimensional FFT is performed.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axes indicated by axes, or the last two axes if axes is not given.
Raises:

ValueError

If s and axes have different length, or axes not given and len(s) != 2.

IndexError

If an element of axes is larger than than the number of axes of a.

See also

numpy.fft
Overall view of discrete Fourier transforms, with definitions and conventions used.
fft2
The forward 2-dimensional FFT, of which ifft2 is the inverse.
ifftn
The inverse of the n-dimensional FFT.
fft
The one-dimensional FFT.
ifft
The one-dimensional inverse FFT.

Notes

ifft2 is just ifftn with a different default for axes.

See ifftn for details and a plotting example, and numpy.fft for definition and conventions used.

Zero-padding, analogously with ifft, is performed by appending zeros to the input along the specified dimension. Although this is the common approach, it might lead to surprising results. If another form of zero padding is desired, it must be performed before ifft2 is called.

Examples

>>> a = 4 * np.eye(4)
>>> np.fft.ifft2(a)
array([[ 1.+0.j,  0.+0.j,  0.+0.j,  0.+0.j],
       [ 0.+0.j,  0.+0.j,  0.+0.j,  1.+0.j],
       [ 0.+0.j,  0.+0.j,  1.+0.j,  0.+0.j],
       [ 0.+0.j,  1.+0.j,  0.+0.j,  0.+0.j]])
dask.array.fft.ifftn(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.ifftn

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.ifftn docstring follows below:

Compute the N-dimensional inverse discrete Fourier Transform.

This function computes the inverse of the N-dimensional discrete Fourier Transform over any number of axes in an M-dimensional array by means of the Fast Fourier Transform (FFT). In other words, ifftn(fftn(a)) == a to within numerical accuracy. For a description of the definitions and conventions used, see numpy.fft.

The input, analogously to ifft, should be ordered in the same way as is returned by fftn, i.e. it should have the term for zero frequency in all axes in the low-order corner, the positive frequency terms in the first half of all axes, the term for the Nyquist frequency in the middle of all axes and the negative frequency terms in the second half of all axes, in order of decreasingly negative frequency.

Parameters:

a : array_like

Input array, can be complex.

s : sequence of ints, optional

Shape (length of each transformed axis) of the output (s[0] refers to axis 0, s[1] to axis 1, etc.). This corresponds to n for ifft(x, n). Along any axis, if the given shape is smaller than that of the input, the input is cropped. If it is larger, the input is padded with zeros. if s is not given, the shape of the input along the axes specified by axes is used. See notes for issue on ifft zero padding.

axes : sequence of ints, optional

Axes over which to compute the IFFT. If not given, the last len(s) axes are used, or all axes if s is also not specified. Repeated indices in axes means that the inverse transform over that axis is performed multiple times.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axes indicated by axes, or by a combination of s or a, as explained in the parameters section above.
Raises:

ValueError

If s and axes have different length.

IndexError

If an element of axes is larger than than the number of axes of a.

See also

numpy.fft
Overall view of discrete Fourier transforms, with definitions and conventions used.
fftn
The forward n-dimensional FFT, of which ifftn is the inverse.
ifft
The one-dimensional inverse FFT.
ifft2
The two-dimensional inverse FFT.
ifftshift
Undoes fftshift, shifts zero-frequency terms to beginning of array.

Notes

See numpy.fft for definitions and conventions used.

Zero-padding, analogously with ifft, is performed by appending zeros to the input along the specified dimension. Although this is the common approach, it might lead to surprising results. If another form of zero padding is desired, it must be performed before ifftn is called.

Examples

>>> a = np.eye(4)
>>> np.fft.ifftn(np.fft.fftn(a, axes=(0,)), axes=(1,))
array([[ 1.+0.j,  0.+0.j,  0.+0.j,  0.+0.j],
       [ 0.+0.j,  1.+0.j,  0.+0.j,  0.+0.j],
       [ 0.+0.j,  0.+0.j,  1.+0.j,  0.+0.j],
       [ 0.+0.j,  0.+0.j,  0.+0.j,  1.+0.j]])

Create and plot an image with band-limited frequency content:

>>> import matplotlib.pyplot as plt
>>> n = np.zeros((200,200), dtype=complex)
>>> n[60:80, 20:40] = np.exp(1j*np.random.uniform(0, 2*np.pi, (20, 20)))
>>> im = np.fft.ifftn(n).real
>>> plt.imshow(im)
<matplotlib.image.AxesImage object at 0x...>
>>> plt.show()
dask.array.fft.rfft(a, n=None, axis=None)

Wrapping of numpy.fft.fftpack.rfft

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.rfft docstring follows below:

Compute the one-dimensional discrete Fourier Transform for real input.

This function computes the one-dimensional n-point discrete Fourier Transform (DFT) of a real-valued array by means of an efficient algorithm called the Fast Fourier Transform (FFT).

Parameters:

a : array_like

Input array

n : int, optional

Number of points along transformation axis in the input to use. If n is smaller than the length of the input, the input is cropped. If it is larger, the input is padded with zeros. If n is not given, the length of the input along the axis specified by axis is used.

axis : int, optional

Axis over which to compute the FFT. If not given, the last axis is used.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axis indicated by axis, or the last one if axis is not specified. If n is even, the length of the transformed axis is (n/2)+1. If n is odd, the length is (n+1)/2.
Raises:

IndexError

If axis is larger than the last axis of a.

See also

numpy.fft
For definition of the DFT and conventions used.
irfft
The inverse of rfft.
fft
The one-dimensional FFT of general (complex) input.
fftn
The n-dimensional FFT.
rfftn
The n-dimensional FFT of real input.

Notes

When the DFT is computed for purely real input, the output is Hermitian-symmetric, i.e. the negative frequency terms are just the complex conjugates of the corresponding positive-frequency terms, and the negative-frequency terms are therefore redundant. This function does not compute the negative frequency terms, and the length of the transformed axis of the output is therefore n//2 + 1.

When A = rfft(a) and fs is the sampling frequency, A[0] contains the zero-frequency term 0*fs, which is real due to Hermitian symmetry.

If n is even, A[-1] contains the term representing both positive and negative Nyquist frequency (+fs/2 and -fs/2), and must also be purely real. If n is odd, there is no term at fs/2; A[-1] contains the largest positive frequency (fs/2*(n-1)/n), and is complex in the general case.

If the input a contains an imaginary part, it is silently discarded.

Examples

>>> np.fft.fft([0, 1, 0, 0])
array([ 1.+0.j,  0.-1.j, -1.+0.j,  0.+1.j])
>>> np.fft.rfft([0, 1, 0, 0])
array([ 1.+0.j,  0.-1.j, -1.+0.j])

Notice how the final element of the fft output is the complex conjugate of the second element, for real input. For rfft, this symmetry is exploited to compute only the non-negative frequency terms.

dask.array.fft.rfft2(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.rfft2

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.rfft2 docstring follows below:

Compute the 2-dimensional FFT of a real array.

Parameters:

a : array

Input array, taken to be real.

s : sequence of ints, optional

Shape of the FFT.

axes : sequence of ints, optional

Axes over which to compute the FFT.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : ndarray

The result of the real 2-D FFT.

See also

rfftn
Compute the N-dimensional discrete Fourier Transform for real input.

Notes

This is really just rfftn with different default behavior. For more details see rfftn.

dask.array.fft.rfftn(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.rfftn

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.rfftn docstring follows below:

Compute the N-dimensional discrete Fourier Transform for real input.

This function computes the N-dimensional discrete Fourier Transform over any number of axes in an M-dimensional real array by means of the Fast Fourier Transform (FFT). By default, all axes are transformed, with the real transform performed over the last axis, while the remaining transforms are complex.

Parameters:

a : array_like

Input array, taken to be real.

s : sequence of ints, optional

Shape (length along each transformed axis) to use from the input. (s[0] refers to axis 0, s[1] to axis 1, etc.). The final element of s corresponds to n for rfft(x, n), while for the remaining axes, it corresponds to n for fft(x, n). Along any axis, if the given shape is smaller than that of the input, the input is cropped. If it is larger, the input is padded with zeros. if s is not given, the shape of the input along the axes specified by axes is used.

axes : sequence of ints, optional

Axes over which to compute the FFT. If not given, the last len(s) axes are used, or all axes if s is also not specified.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axes indicated by axes, or by a combination of s and a, as explained in the parameters section above. The length of the last axis transformed will be s[-1]//2+1, while the remaining transformed axes will have lengths according to s, or unchanged from the input.
Raises:

ValueError

If s and axes have different length.

IndexError

If an element of axes is larger than than the number of axes of a.

See also

irfftn
The inverse of rfftn, i.e. the inverse of the n-dimensional FFT of real input.
fft
The one-dimensional FFT, with definitions and conventions used.
rfft
The one-dimensional FFT of real input.
fftn
The n-dimensional FFT.
rfft2
The two-dimensional FFT of real input.

Notes

The transform for real input is performed over the last transformation axis, as by rfft, then the transform over the remaining axes is performed as by fftn. The order of the output is as for rfft for the final transformation axis, and as for fftn for the remaining transformation axes.

See fft for details, definitions and conventions used.

Examples

>>> a = np.ones((2, 2, 2))
>>> np.fft.rfftn(a)
array([[[ 8.+0.j,  0.+0.j],
        [ 0.+0.j,  0.+0.j]],
       [[ 0.+0.j,  0.+0.j],
        [ 0.+0.j,  0.+0.j]]])

>>> np.fft.rfftn(a, axes=(2, 0))
array([[[ 4.+0.j,  0.+0.j],
        [ 4.+0.j,  0.+0.j]],
       [[ 0.+0.j,  0.+0.j],
        [ 0.+0.j,  0.+0.j]]])
dask.array.fft.irfft(a, n=None, axis=None)

Wrapping of numpy.fft.fftpack.irfft

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.irfft docstring follows below:

Compute the inverse of the n-point DFT for real input.

This function computes the inverse of the one-dimensional n-point discrete Fourier Transform of real input computed by rfft. In other words, irfft(rfft(a), len(a)) == a to within numerical accuracy. (See Notes below for why len(a) is necessary here.)

The input is expected to be in the form returned by rfft, i.e. the real zero-frequency term followed by the complex positive frequency terms in order of increasing frequency. Since the discrete Fourier Transform of real input is Hermitian-symmetric, the negative frequency terms are taken to be the complex conjugates of the corresponding positive frequency terms.

Parameters:

a : array_like

The input array.

n : int, optional

Length of the transformed axis of the output. For n output points, n//2+1 input points are necessary. If the input is longer than this, it is cropped. If it is shorter than this, it is padded with zeros. If n is not given, it is determined from the length of the input along the axis specified by axis.

axis : int, optional

Axis over which to compute the inverse FFT. If not given, the last axis is used.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : ndarray

The truncated or zero-padded input, transformed along the axis indicated by axis, or the last one if axis is not specified. The length of the transformed axis is n, or, if n is not given, 2*(m-1) where m is the length of the transformed axis of the input. To get an odd number of output points, n must be specified.
Raises:

IndexError

If axis is larger than the last axis of a.

See also

numpy.fft
For definition of the DFT and conventions used.
rfft
The one-dimensional FFT of real input, of which irfft is inverse.
fft
The one-dimensional FFT.
irfft2
The inverse of the two-dimensional FFT of real input.
irfftn
The inverse of the n-dimensional FFT of real input.

Notes

Returns the real valued n-point inverse discrete Fourier transform of a, where a contains the non-negative frequency terms of a Hermitian-symmetric sequence. n is the length of the result, not the input.

If you specify an n such that a must be zero-padded or truncated, the extra/removed values will be added/removed at high frequencies. One can thus resample a series to m points via Fourier interpolation by: a_resamp = irfft(rfft(a), m).

Examples

>>> np.fft.ifft([1, -1j, -1, 1j])
array([ 0.+0.j,  1.+0.j,  0.+0.j,  0.+0.j])
>>> np.fft.irfft([1, -1j, -1])
array([ 0.,  1.,  0.,  0.])

Notice how the last term in the input to the ordinary ifft is the complex conjugate of the second term, and the output has zero imaginary part everywhere. When calling irfft, the negative frequencies are not specified, and the output array is purely real.

dask.array.fft.irfft2(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.irfft2

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.irfft2 docstring follows below:

Compute the 2-dimensional inverse FFT of a real array.

Parameters:

a : array_like

The input array

s : sequence of ints, optional

Shape of the inverse FFT.

axes : sequence of ints, optional

The axes over which to compute the inverse fft. Default is the last two axes.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : ndarray

The result of the inverse real 2-D FFT.

See also

irfftn
Compute the inverse of the N-dimensional FFT of real input.

Notes

This is really irfftn with different defaults. For more details see irfftn.

dask.array.fft.irfftn(a, s=None, axes=None)

Wrapping of numpy.fft.fftpack.irfftn

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.irfftn docstring follows below:

Compute the inverse of the N-dimensional FFT of real input.

This function computes the inverse of the N-dimensional discrete Fourier Transform for real input over any number of axes in an M-dimensional array by means of the Fast Fourier Transform (FFT). In other words, irfftn(rfftn(a), a.shape) == a to within numerical accuracy. (The a.shape is necessary like len(a) is for irfft, and for the same reason.)

The input should be ordered in the same way as is returned by rfftn, i.e. as for irfft for the final transformation axis, and as for ifftn along all the other axes.

Parameters:

a : array_like

Input array.

s : sequence of ints, optional

Shape (length of each transformed axis) of the output (s[0] refers to axis 0, s[1] to axis 1, etc.). s is also the number of input points used along this axis, except for the last axis, where s[-1]//2+1 points of the input are used. Along any axis, if the shape indicated by s is smaller than that of the input, the input is cropped. If it is larger, the input is padded with zeros. If s is not given, the shape of the input along the axes specified by axes is used.

axes : sequence of ints, optional

Axes over which to compute the inverse FFT. If not given, the last len(s) axes are used, or all axes if s is also not specified. Repeated indices in axes means that the inverse transform over that axis is performed multiple times.

norm : {None, “ortho”}, optional

New in version 1.10.0.

Normalization mode (see numpy.fft). Default is None.
Returns:

out : ndarray

The truncated or zero-padded input, transformed along the axes indicated by axes, or by a combination of s or a, as explained in the parameters section above. The length of each transformed axis is as given by the corresponding element of s, or the length of the input in every axis except for the last one if s is not given. In the final transformed axis the length of the output when s is not given is 2*(m-1) where m is the length of the final transformed axis of the input. To get an odd number of output points in the final axis, s must be specified.
Raises:

ValueError

If s and axes have different length.

IndexError

If an element of axes is larger than than the number of axes of a.

See also

rfftn
The forward n-dimensional FFT of real input, of which ifftn is the inverse.
fft
The one-dimensional FFT, with definitions and conventions used.
irfft
The inverse of the one-dimensional FFT of real input.
irfft2
The inverse of the two-dimensional FFT of real input.

Notes

See fft for definitions and conventions used.

See rfft for definitions and conventions used for real input.

Examples

>>> a = np.zeros((3, 2, 2))
>>> a[0, 0, 0] = 3 * 2 * 2
>>> np.fft.irfftn(a)
array([[[ 1.,  1.],
        [ 1.,  1.]],
       [[ 1.,  1.],
        [ 1.,  1.]],
       [[ 1.,  1.],
        [ 1.,  1.]]])
dask.array.fft.hfft(a, n=None, axis=None)

Wrapping of numpy.fft.fftpack.hfft

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.hfft docstring follows below:

Compute the FFT of a signal that has Hermitian symmetry, i.e., a real spectrum.

Parameters:

a : array_like

The input array.

n : int, optional

Length of the transformed axis of the output. For n output points, n//2 + 1 input points are necessary. If the input is longer than this, it is cropped. If it is shorter than this, it is padded with zeros. If n is not given, it is determined from the length of the input along the axis specified by axis.

axis : int, optional

Axis over which to compute the FFT. If not given, the last axis is used.

norm : {None, “ortho”}, optional

Normalization mode (see numpy.fft). Default is None.

New in version 1.10.0.
Returns:

out : ndarray

The truncated or zero-padded input, transformed along the axis indicated by axis, or the last one if axis is not specified. The length of the transformed axis is n, or, if n is not given, 2*m - 2 where m is the length of the transformed axis of the input. To get an odd number of output points, n must be specified, for instance as 2*m - 1 in the typical case,
Raises:

IndexError

If axis is larger than the last axis of a.

See also

rfft
Compute the one-dimensional FFT for real input.
ihfft
The inverse of hfft.

Notes

hfft/ihfft are a pair analogous to rfft/irfft, but for the opposite case: here the signal has Hermitian symmetry in the time domain and is real in the frequency domain. So here it’s hfft for which you must supply the length of the result if it is to be odd.

  • even: ihfft(hfft(a, 2*len(a) - 2) == a, within roundoff error,
  • odd: ihfft(hfft(a, 2*len(a) - 1) == a, within roundoff error.

Examples

>>> signal = np.array([1, 2, 3, 4, 3, 2])
>>> np.fft.fft(signal)
array([ 15.+0.j,  -4.+0.j,   0.+0.j,  -1.-0.j,   0.+0.j,  -4.+0.j])
>>> np.fft.hfft(signal[:4]) # Input first half of signal
array([ 15.,  -4.,   0.,  -1.,   0.,  -4.])
>>> np.fft.hfft(signal, 6)  # Input entire signal and truncate
array([ 15.,  -4.,   0.,  -1.,   0.,  -4.])

>>> signal = np.array([[1, 1.j], [-1.j, 2]])
>>> np.conj(signal.T) - signal   # check Hermitian symmetry
array([[ 0.-0.j,  0.+0.j],
       [ 0.+0.j,  0.-0.j]])
>>> freq_spectrum = np.fft.hfft(signal)
>>> freq_spectrum
array([[ 1.,  1.],
       [ 2., -2.]])
dask.array.fft.ihfft(a, n=None, axis=None)

Wrapping of numpy.fft.fftpack.ihfft

The axis along which the FFT is applied must have a one chunk. To change the array’s chunking use dask.Array.rechunk.

The numpy.fft.fftpack.ihfft docstring follows below:

Compute the inverse FFT of a signal that has Hermitian symmetry.

Parameters:

a : array_like

Input array.

n : int, optional

Length of the inverse FFT, the number of points along transformation axis in the input to use. If n is smaller than the length of the input, the input is cropped. If it is larger, the input is padded with zeros. If n is not given, the length of the input along the axis specified by axis is used.

axis : int, optional

Axis over which to compute the inverse FFT. If not given, the last axis is used.

norm : {None, “ortho”}, optional

Normalization mode (see numpy.fft). Default is None.

New in version 1.10.0.
Returns:

out : complex ndarray

The truncated or zero-padded input, transformed along the axis indicated by axis, or the last one if axis is not specified. The length of the transformed axis is n//2 + 1.

See also

hfft, irfft

Notes

hfft/ihfft are a pair analogous to rfft/irfft, but for the opposite case: here the signal has Hermitian symmetry in the time domain and is real in the frequency domain. So here it’s hfft for which you must supply the length of the result if it is to be odd:

  • even: ihfft(hfft(a, 2*len(a) - 2) == a, within roundoff error,
  • odd: ihfft(hfft(a, 2*len(a) - 1) == a, within roundoff error.

Examples

>>> spectrum = np.array([ 15, -4, 0, -1, 0, -4])
>>> np.fft.ifft(spectrum)
array([ 1.+0.j,  2.-0.j,  3.+0.j,  4.+0.j,  3.+0.j,  2.-0.j])
>>> np.fft.ihfft(spectrum)
array([ 1.-0.j,  2.-0.j,  3.-0.j,  4.-0.j])
dask.array.fft.fftfreq(n, d=1.0)

Return the Discrete Fourier Transform sample frequencies.

The returned float array f contains the frequency bin centers in cycles per unit of the sample spacing (with zero at the start). For instance, if the sample spacing is in seconds, then the frequency unit is cycles/second.

Given a window length n and a sample spacing d:

f = [0, 1, ...,   n/2-1,     -n/2, ..., -1] / (d*n)   if n is even
f = [0, 1, ..., (n-1)/2, -(n-1)/2, ..., -1] / (d*n)   if n is odd
Parameters:

n : int

Window length.

d : scalar, optional

Sample spacing (inverse of the sampling rate). Defaults to 1.
Returns:

f : ndarray

Array of length n containing the sample frequencies.

Examples

>>> signal = np.array([-2, 8, 6, 4, 1, 0, 3, 5], dtype=float)
>>> fourier = np.fft.fft(signal)
>>> n = signal.size
>>> timestep = 0.1
>>> freq = np.fft.fftfreq(n, d=timestep)
>>> freq
array([ 0.  ,  1.25,  2.5 ,  3.75, -5.  , -3.75, -2.5 , -1.25])
dask.array.fft.rfftfreq(n, d=1.0)

Return the Discrete Fourier Transform sample frequencies (for usage with rfft, irfft).

The returned float array f contains the frequency bin centers in cycles per unit of the sample spacing (with zero at the start). For instance, if the sample spacing is in seconds, then the frequency unit is cycles/second.

Given a window length n and a sample spacing d:

f = [0, 1, ...,     n/2-1,     n/2] / (d*n)   if n is even
f = [0, 1, ..., (n-1)/2-1, (n-1)/2] / (d*n)   if n is odd

Unlike fftfreq (but like scipy.fftpack.rfftfreq) the Nyquist frequency component is considered to be positive.

Parameters:

n : int

Window length.

d : scalar, optional

Sample spacing (inverse of the sampling rate). Defaults to 1.
Returns:

f : ndarray

Array of length n//2 + 1 containing the sample frequencies.

Examples

>>> signal = np.array([-2, 8, 6, 4, 1, 0, 3, 5, -3, 4], dtype=float)
>>> fourier = np.fft.rfft(signal)
>>> n = signal.size
>>> sample_rate = 100
>>> freq = np.fft.fftfreq(n, d=1./sample_rate)
>>> freq
array([  0.,  10.,  20.,  30.,  40., -50., -40., -30., -20., -10.])
>>> freq = np.fft.rfftfreq(n, d=1./sample_rate)
>>> freq
array([  0.,  10.,  20.,  30.,  40.,  50.])
dask.array.fft.fftshift(x, axes=None)

Shift the zero-frequency component to the center of the spectrum.

This function swaps half-spaces for all axes listed (defaults to all). Note that y[0] is the Nyquist component only if len(x) is even.

Parameters:

x : array_like

Input array.

axes : int or shape tuple, optional

Axes over which to shift. Default is None, which shifts all axes.
Returns:

y : ndarray

The shifted array.

See also

ifftshift
The inverse of fftshift.

Examples

>>> freqs = np.fft.fftfreq(10, 0.1)
>>> freqs
array([ 0.,  1.,  2.,  3.,  4., -5., -4., -3., -2., -1.])
>>> np.fft.fftshift(freqs)
array([-5., -4., -3., -2., -1.,  0.,  1.,  2.,  3.,  4.])

Shift the zero-frequency component only along the second axis:

>>> freqs = np.fft.fftfreq(9, d=1./9).reshape(3, 3)
>>> freqs
array([[ 0.,  1.,  2.],
       [ 3.,  4., -4.],
       [-3., -2., -1.]])
>>> np.fft.fftshift(freqs, axes=(1,))
array([[ 2.,  0.,  1.],
       [-4.,  3.,  4.],
       [-1., -3., -2.]])
dask.array.fft.ifftshift(x, axes=None)

The inverse of fftshift. Although identical for even-length x, the functions differ by one sample for odd-length x.

Parameters:

x : array_like

Input array.

axes : int or shape tuple, optional

Axes over which to calculate. Defaults to None, which shifts all axes.
Returns:

y : ndarray

The shifted array.

See also

fftshift
Shift zero-frequency component to the center of the spectrum.

Examples

>>> freqs = np.fft.fftfreq(9, d=1./9).reshape(3, 3)
>>> freqs
array([[ 0.,  1.,  2.],
       [ 3.,  4., -4.],
       [-3., -2., -1.]])
>>> np.fft.ifftshift(np.fft.fftshift(freqs))
array([[ 0.,  1.,  2.],
       [ 3.,  4., -4.],
       [-3., -2., -1.]])
dask.array.random.beta(a, b, size=None)

Draw samples from a Beta distribution.

The Beta distribution is a special case of the Dirichlet distribution, and is related to the Gamma distribution. It has the probability distribution function

\[f(x; a,b) = \frac{1}{B(\alpha, \beta)} x^{\alpha - 1} (1 - x)^{\beta - 1},\]

where the normalisation, B, is the beta function,

\[B(\alpha, \beta) = \int_0^1 t^{\alpha - 1} (1 - t)^{\beta - 1} dt.\]

It is often seen in Bayesian inference and order statistics.

Parameters:

a : float or array_like of floats

Alpha, non-negative.

b : float or array_like of floats

Beta, non-negative.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if a and b are both scalars. Otherwise, np.broadcast(a, b).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized beta distribution.
dask.array.random.binomial(n, p, size=None)

Draw samples from a binomial distribution.

Samples are drawn from a binomial distribution with specified parameters, n trials and p probability of success where n an integer >= 0 and p is in the interval [0,1]. (n may be input as a float, but it is truncated to an integer in use)

Parameters:

n : int or array_like of ints

Parameter of the distribution, >= 0. Floats are also accepted, but they will be truncated to integers.

p : float or array_like of floats

Parameter of the distribution, >= 0 and <=1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if n and p are both scalars. Otherwise, np.broadcast(n, p).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized binomial distribution, where each sample is equal to the number of successes over the n trials.

See also

scipy.stats.binom
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the binomial distribution is

\[P(N) = \binom{n}{N}p^N(1-p)^{n-N},\]

where \(n\) is the number of trials, \(p\) is the probability of success, and \(N\) is the number of successes.

When estimating the standard error of a proportion in a population by using a random sample, the normal distribution works well unless the product p*n <=5, where p = population proportion estimate, and n = number of samples, in which case the binomial distribution is used instead. For example, a sample of 15 people shows 4 who are left handed, and 11 who are right handed. Then p = 4/15 = 27%. 0.27*15 = 4, so the binomial distribution should be used in this case.

References

[R247251]Dalgaard, Peter, “Introductory Statistics with R”, Springer-Verlag, 2002.
[R248251]Glantz, Stanton A. “Primer of Biostatistics.”, McGraw-Hill, Fifth Edition, 2002.
[R249251]Lentner, Marvin, “Elementary Applied Statistics”, Bogden and Quigley, 1972.
[R250251]Weisstein, Eric W. “Binomial Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/BinomialDistribution.html
[R251251]Wikipedia, “Binomial distribution”, http://en.wikipedia.org/wiki/Binomial_distribution

Examples

Draw samples from the distribution:

>> n, p = 10, .5 # number of trials, probability of each trial >> s = np.random.binomial(n, p, 1000) # result of flipping a coin 10 times, tested 1000 times.

A real world example. A company drills 9 wild-cat oil exploration wells, each with an estimated probability of success of 0.1. All nine wells fail. What is the probability of that happening?

Let’s do 20,000 trials of the model, and count the number that generate zero positive results.

>> sum(np.random.binomial(9, 0.1, 20000) == 0)/20000. # answer = 0.38885, or 38%.

dask.array.random.chisquare(df, size=None)

Draw samples from a chi-square distribution.

When df independent random variables, each with standard normal distributions (mean 0, variance 1), are squared and summed, the resulting distribution is chi-square (see Notes). This distribution is often used in hypothesis testing.

Parameters:

df : int or array_like of ints

Number of degrees of freedom.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if df is a scalar. Otherwise, np.array(df).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized chi-square distribution.
Raises:

ValueError

When df <= 0 or when an inappropriate size (e.g. size=-1) is given.

Notes

The variable obtained by summing the squares of df independent, standard normally distributed random variables:

\[Q = \sum_{i=0}^{\mathtt{df}} X^2_i\]

is chi-square distributed, denoted

\[Q \sim \chi^2_k.\]

The probability density function of the chi-squared distribution is

\[p(x) = \frac{(1/2)^{k/2}}{\Gamma(k/2)} x^{k/2 - 1} e^{-x/2},\]

where \(\Gamma\) is the gamma function,

\[\Gamma(x) = \int_0^{-\infty} t^{x - 1} e^{-t} dt.\]

References

[R257257]NIST “Engineering Statistics Handbook” http://www.itl.nist.gov/div898/handbook/eda/section3/eda3666.htm

Examples

>> np.random.chisquare(2,4) array([ 1.89920014, 9.00867716, 3.13710533, 5.62318272])

dask.array.random.exponential(scale=1.0, size=None)

Draw samples from an exponential distribution.

Its probability density function is

\[f(x; \frac{1}{\beta}) = \frac{1}{\beta} \exp(-\frac{x}{\beta}),\]

for x > 0 and 0 elsewhere. \(\beta\) is the scale parameter, which is the inverse of the rate parameter \(\lambda = 1/\beta\). The rate parameter is an alternative, widely used parameterization of the exponential distribution [R261261].

The exponential distribution is a continuous analogue of the geometric distribution. It describes many common situations, such as the size of raindrops measured over many rainstorms [R259261], or the time between page requests to Wikipedia [R260261].

Parameters:

scale : float or array_like of floats

The scale parameter, \(\beta = 1/\lambda\).

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if scale is a scalar. Otherwise, np.array(scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized exponential distribution.

References

[R259261](1, 2) Peyton Z. Peebles Jr., “Probability, Random Variables and Random Signal Principles”, 4th ed, 2001, p. 57.
[R260261](1, 2) Wikipedia, “Poisson process”, http://en.wikipedia.org/wiki/Poisson_process
[R261261](1, 2) Wikipedia, “Exponential distribution”, http://en.wikipedia.org/wiki/Exponential_distribution
dask.array.random.f(dfnum, dfden, size=None)

Draw samples from an F distribution.

Samples are drawn from an F distribution with specified parameters, dfnum (degrees of freedom in numerator) and dfden (degrees of freedom in denominator), where both parameters should be greater than zero.

The random variate of the F distribution (also known as the Fisher distribution) is a continuous probability distribution that arises in ANOVA tests, and is the ratio of two chi-square variates.

Parameters:

dfnum : int or array_like of ints

Degrees of freedom in numerator. Should be greater than zero.

dfden : int or array_like of ints

Degrees of freedom in denominator. Should be greater than zero.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if dfnum and dfden are both scalars. Otherwise, np.broadcast(dfnum, dfden).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Fisher distribution.

See also

scipy.stats.f
probability density function, distribution or cumulative density function, etc.

Notes

The F statistic is used to compare in-group variances to between-group variances. Calculating the distribution depends on the sampling, and so it is a function of the respective degrees of freedom in the problem. The variable dfnum is the number of samples minus one, the between-groups degrees of freedom, while dfden is the within-groups degrees of freedom, the sum of the number of samples in each group minus the number of groups.

References

[R265266]Glantz, Stanton A. “Primer of Biostatistics.”, McGraw-Hill, Fifth Edition, 2002.
[R266266]Wikipedia, “F-distribution”, http://en.wikipedia.org/wiki/F-distribution

Examples

An example from Glantz[1], pp 47-40:

Two groups, children of diabetics (25 people) and children from people without diabetes (25 controls). Fasting blood glucose was measured, case group had a mean value of 86.1, controls had a mean value of 82.2. Standard deviations were 2.09 and 2.49 respectively. Are these data consistent with the null hypothesis that the parents diabetic status does not affect their children’s blood glucose levels? Calculating the F statistic from the data gives a value of 36.01.

Draw samples from the distribution:

>> dfnum = 1. # between group degrees of freedom >> dfden = 48. # within groups degrees of freedom >> s = np.random.f(dfnum, dfden, 1000)

The lower bound for the top 1% of the samples is :

>> sort(s)[-10] 7.61988120985

So there is about a 1% chance that the F statistic will exceed 7.62, the measured value is 36, so the null hypothesis is rejected at the 1% level.

dask.array.random.gamma(shape, scale=1.0, size=None)

Draw samples from a Gamma distribution.

Samples are drawn from a Gamma distribution with specified parameters, shape (sometimes designated “k”) and scale (sometimes designated “theta”), where both parameters are > 0.

Parameters:

shape : float or array_like of floats

The shape of the gamma distribution. Should be greater than zero.

scale : float or array_like of floats, optional

The scale of the gamma distribution. Should be greater than zero. Default is equal to 1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if shape and scale are both scalars. Otherwise, np.broadcast(shape, scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized gamma distribution.

See also

scipy.stats.gamma
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Gamma distribution is

\[p(x) = x^{k-1}\frac{e^{-x/\theta}}{\theta^k\Gamma(k)},\]

where \(k\) is the shape and \(\theta\) the scale, and \(\Gamma\) is the Gamma function.

The Gamma distribution is often used to model the times to failure of electronic components, and arises naturally in processes for which the waiting times between Poisson distributed events are relevant.

References

[R269270]Weisstein, Eric W. “Gamma Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/GammaDistribution.html
[R270270]Wikipedia, “Gamma distribution”, http://en.wikipedia.org/wiki/Gamma_distribution

Examples

Draw samples from the distribution:

>> shape, scale = 2., 2. # mean=4, std=2*sqrt(2) >> s = np.random.gamma(shape, scale, 1000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> import scipy.special as sps >> count, bins, ignored = plt.hist(s, 50, normed=True) >> y = bins**(shape-1)*(np.exp(-bins/scale) / .. (sps.gamma(shape)*scale**shape)) >> plt.plot(bins, y, linewidth=2, color=’r’) >> plt.show()

dask.array.random.geometric(p, size=None)

Draw samples from the geometric distribution.

Bernoulli trials are experiments with one of two outcomes: success or failure (an example of such an experiment is flipping a coin). The geometric distribution models the number of trials that must be run in order to achieve success. It is therefore supported on the positive integers, k = 1, 2, ...

The probability mass function of the geometric distribution is

\[f(k) = (1 - p)^{k - 1} p\]

where p is the probability of success of an individual trial.

Parameters:

p : float or array_like of floats

The probability of success of an individual trial.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if p is a scalar. Otherwise, np.array(p).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized geometric distribution.

Examples

Draw ten thousand values from the geometric distribution, with the probability of an individual success equal to 0.35:

>> z = np.random.geometric(p=0.35, size=10000)

How many trials succeeded after a single run?

>> (z == 1).sum() / 10000. 0.34889999999999999 #random

dask.array.random.gumbel(loc=0.0, scale=1.0, size=None)

Draw samples from a Gumbel distribution.

Draw samples from a Gumbel distribution with specified location and scale. For more information on the Gumbel distribution, see Notes and References below.

Parameters:

loc : float or array_like of floats, optional

The location of the mode of the distribution. Default is 0.

scale : float or array_like of floats, optional

The scale parameter of the distribution. Default is 1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if loc and scale are both scalars. Otherwise, np.broadcast(loc, scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Gumbel distribution.

See also

scipy.stats.gumbel_l, scipy.stats.gumbel_r, scipy.stats.genextreme, weibull

Notes

The Gumbel (or Smallest Extreme Value (SEV) or the Smallest Extreme Value Type I) distribution is one of a class of Generalized Extreme Value (GEV) distributions used in modeling extreme value problems. The Gumbel is a special case of the Extreme Value Type I distribution for maximums from distributions with “exponential-like” tails.

The probability density for the Gumbel distribution is

\[p(x) = \frac{e^{-(x - \mu)/ \beta}}{\beta} e^{ -e^{-(x - \mu)/ \beta}},\]

where \(\mu\) is the mode, a location parameter, and \(\beta\) is the scale parameter.

The Gumbel (named for German mathematician Emil Julius Gumbel) was used very early in the hydrology literature, for modeling the occurrence of flood events. It is also used for modeling maximum wind speed and rainfall rates. It is a “fat-tailed” distribution - the probability of an event in the tail of the distribution is larger than if one used a Gaussian, hence the surprisingly frequent occurrence of 100-year floods. Floods were initially modeled as a Gaussian process, which underestimated the frequency of extreme events.

It is one of a class of extreme value distributions, the Generalized Extreme Value (GEV) distributions, which also includes the Weibull and Frechet.

The function has a mean of \(\mu + 0.57721\beta\) and a variance of \(\frac{\pi^2}{6}\beta^2\).

References

[R273274]Gumbel, E. J., “Statistics of Extremes,” New York: Columbia University Press, 1958.
[R274274]Reiss, R.-D. and Thomas, M., “Statistical Analysis of Extreme Values from Insurance, Finance, Hydrology and Other Fields,” Basel: Birkhauser Verlag, 2001.

Examples

Draw samples from the distribution:

>> mu, beta = 0, 0.1 # location and scale >> s = np.random.gumbel(mu, beta, 1000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, 30, normed=True) >> plt.plot(bins, (1/beta)*np.exp(-(bins - mu)/beta) .. * np.exp( -np.exp( -(bins - mu) /beta) ), .. linewidth=2, color=’r’) >> plt.show()

Show how an extreme value distribution can arise from a Gaussian process and compare to a Gaussian:

>> means = [] >> maxima = [] >> for i in range(0,1000) : .. a = np.random.normal(mu, beta, 1000) .. means.append(a.mean()) .. maxima.append(a.max()) >> count, bins, ignored = plt.hist(maxima, 30, normed=True) >> beta = np.std(maxima) * np.sqrt(6) / np.pi >> mu = np.mean(maxima) - 0.57721*beta >> plt.plot(bins, (1/beta)*np.exp(-(bins - mu)/beta) .. * np.exp(-np.exp(-(bins - mu)/beta)), .. linewidth=2, color=’r’) >> plt.plot(bins, 1/(beta * np.sqrt(2 * np.pi)) .. * np.exp(-(bins - mu)**2 / (2 * beta**2)), .. linewidth=2, color=’g’) >> plt.show()

dask.array.random.hypergeometric(ngood, nbad, nsample, size=None)

Draw samples from a Hypergeometric distribution.

Samples are drawn from a hypergeometric distribution with specified parameters, ngood (ways to make a good selection), nbad (ways to make a bad selection), and nsample = number of items sampled, which is less than or equal to the sum ngood + nbad.

Parameters:

ngood : int or array_like of ints

Number of ways to make a good selection. Must be nonnegative.

nbad : int or array_like of ints

Number of ways to make a bad selection. Must be nonnegative.

nsample : int or array_like of ints

Number of items sampled. Must be at least 1 and at most ngood + nbad.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if ngood, nbad, and nsample are all scalars. Otherwise, np.broadcast(ngood, nbad, nsample).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized hypergeometric distribution.

See also

scipy.stats.hypergeom
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Hypergeometric distribution is

\[P(x) = \frac{\binom{m}{n}\binom{N-m}{n-x}}{\binom{N}{n}},\]

where \(0 \le x \le m\) and \(n+m-N \le x \le n\)

for P(x) the probability of x successes, n = ngood, m = nbad, and N = number of samples.

Consider an urn with black and white marbles in it, ngood of them black and nbad are white. If you draw nsample balls without replacement, then the hypergeometric distribution describes the distribution of black balls in the drawn sample.

Note that this distribution is very similar to the binomial distribution, except that in this case, samples are drawn without replacement, whereas in the Binomial case samples are drawn with replacement (or the sample space is infinite). As the sample space becomes large, this distribution approaches the binomial.

References

[R277279]Lentner, Marvin, “Elementary Applied Statistics”, Bogden and Quigley, 1972.
[R278279]Weisstein, Eric W. “Hypergeometric Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/HypergeometricDistribution.html
[R279279]Wikipedia, “Hypergeometric distribution”, http://en.wikipedia.org/wiki/Hypergeometric_distribution

Examples

Draw samples from the distribution:

>> ngood, nbad, nsamp = 100, 2, 10 # number of good, number of bad, and number of samples >> s = np.random.hypergeometric(ngood, nbad, nsamp, 1000) >> hist(s) # note that it is very unlikely to grab both bad items

Suppose you have an urn with 15 white and 15 black marbles. If you pull 15 marbles at random, how likely is it that 12 or more of them are one color?

>> s = np.random.hypergeometric(15, 15, 15, 100000) >> sum(s>=12)/100000. + sum(s<=3)/100000. # answer = 0.003 .. pretty unlikely!

dask.array.random.laplace(loc=0.0, scale=1.0, size=None)

Draw samples from the Laplace or double exponential distribution with specified location (or mean) and scale (decay).

The Laplace distribution is similar to the Gaussian/normal distribution, but is sharper at the peak and has fatter tails. It represents the difference between two independent, identically distributed exponential random variables.

Parameters:

loc : float or array_like of floats, optional

The position, \(\mu\), of the distribution peak. Default is 0.

scale : float or array_like of floats, optional

\(\lambda\), the exponential decay. Default is 1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if loc and scale are both scalars. Otherwise, np.broadcast(loc, scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Laplace distribution.

Notes

It has the probability density function

\[f(x; \mu, \lambda) = \frac{1}{2\lambda} \exp\left(-\frac{|x - \mu|}{\lambda}\right).\]

The first law of Laplace, from 1774, states that the frequency of an error can be expressed as an exponential function of the absolute magnitude of the error, which leads to the Laplace distribution. For many problems in economics and health sciences, this distribution seems to model the data better than the standard Gaussian distribution.

References

[R283286]Abramowitz, M. and Stegun, I. A. (Eds.). “Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, 9th printing,” New York: Dover, 1972.
[R284286]Kotz, Samuel, et. al. “The Laplace Distribution and Generalizations, ” Birkhauser, 2001.
[R285286]Weisstein, Eric W. “Laplace Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/LaplaceDistribution.html
[R286286]Wikipedia, “Laplace distribution”, http://en.wikipedia.org/wiki/Laplace_distribution

Examples

Draw samples from the distribution

>> loc, scale = 0., 1. >> s = np.random.laplace(loc, scale, 1000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, 30, normed=True) >> x = np.arange(-8., 8., .01) >> pdf = np.exp(-abs(x-loc)/scale)/(2.*scale) >> plt.plot(x, pdf)

Plot Gaussian for comparison:

>> g = (1/(scale * np.sqrt(2 * np.pi)) * .. np.exp(-(x - loc)**2 / (2 * scale**2))) >> plt.plot(x,g)

dask.array.random.logistic(loc=0.0, scale=1.0, size=None)

Draw samples from a logistic distribution.

Samples are drawn from a logistic distribution with specified parameters, loc (location or mean, also median), and scale (>0).

Parameters:

loc : float or array_like of floats, optional

Parameter of the distribution. Default is 0.

scale : float or array_like of floats, optional

Parameter of the distribution. Should be greater than zero. Default is 1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if loc and scale are both scalars. Otherwise, np.broadcast(loc, scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized logistic distribution.

See also

scipy.stats.logistic
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Logistic distribution is

\[P(x) = P(x) = \frac{e^{-(x-\mu)/s}}{s(1+e^{-(x-\mu)/s})^2},\]

where \(\mu\) = location and \(s\) = scale.

The Logistic distribution is used in Extreme Value problems where it can act as a mixture of Gumbel distributions, in Epidemiology, and by the World Chess Federation (FIDE) where it is used in the Elo ranking system, assuming the performance of each player is a logistically distributed random variable.

References

[R291293]Reiss, R.-D. and Thomas M. (2001), “Statistical Analysis of Extreme Values, from Insurance, Finance, Hydrology and Other Fields,” Birkhauser Verlag, Basel, pp 132-133.
[R292293]Weisstein, Eric W. “Logistic Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/LogisticDistribution.html
[R293293]Wikipedia, “Logistic-distribution”, http://en.wikipedia.org/wiki/Logistic_distribution

Examples

Draw samples from the distribution:

>> loc, scale = 10, 1 >> s = np.random.logistic(loc, scale, 10000) >> count, bins, ignored = plt.hist(s, bins=50)

# plot against distribution

>> def logist(x, loc, scale): .. return exp((loc-x)/scale)/(scale*(1+exp((loc-x)/scale))**2) >> plt.plot(bins, logist(bins, loc, scale)*count.max()/.. logist(bins, loc, scale).max()) >> plt.show()

dask.array.random.lognormal(mean=0.0, sigma=1.0, size=None)

Draw samples from a log-normal distribution.

Draw samples from a log-normal distribution with specified mean, standard deviation, and array shape. Note that the mean and standard deviation are not the values for the distribution itself, but of the underlying normal distribution it is derived from.

Parameters:

mean : float or array_like of floats, optional

Mean value of the underlying normal distribution. Default is 0.

sigma : float or array_like of floats, optional

Standard deviation of the underlying normal distribution. Should be greater than zero. Default is 1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if mean and sigma are both scalars. Otherwise, np.broadcast(mean, sigma).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized log-normal distribution.

See also

scipy.stats.lognorm
probability density function, distribution, cumulative density function, etc.

Notes

A variable x has a log-normal distribution if log(x) is normally distributed. The probability density function for the log-normal distribution is:

\[p(x) = \frac{1}{\sigma x \sqrt{2\pi}} e^{(-\frac{(ln(x)-\mu)^2}{2\sigma^2})}\]

where \(\mu\) is the mean and \(\sigma\) is the standard deviation of the normally distributed logarithm of the variable. A log-normal distribution results if a random variable is the product of a large number of independent, identically-distributed variables in the same way that a normal distribution results if the variable is the sum of a large number of independent, identically-distributed variables.

References

[R297298]Limpert, E., Stahel, W. A., and Abbt, M., “Log-normal Distributions across the Sciences: Keys and Clues,” BioScience, Vol. 51, No. 5, May, 2001. http://stat.ethz.ch/~stahel/lognormal/bioscience.pdf
[R298298]Reiss, R.D. and Thomas, M., “Statistical Analysis of Extreme Values,” Basel: Birkhauser Verlag, 2001, pp. 31-32.

Examples

Draw samples from the distribution:

>> mu, sigma = 3., 1. # mean and standard deviation >> s = np.random.lognormal(mu, sigma, 1000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, 100, normed=True, align=’mid’)

>> x = np.linspace(min(bins), max(bins), 10000) >> pdf = (np.exp(-(np.log(x) - mu)**2 / (2 * sigma**2)) .. / (x * sigma * np.sqrt(2 * np.pi)))

>> plt.plot(x, pdf, linewidth=2, color=’r’) >> plt.axis(‘tight’) >> plt.show()

Demonstrate that taking the products of random samples from a uniform distribution can be fit well by a log-normal probability density function.

>> # Generate a thousand samples: each is the product of 100 random >> # values, drawn from a normal distribution. >> b = [] >> for i in range(1000): .. a = 10. + np.random.random(100) .. b.append(np.product(a))

>> b = np.array(b) / np.min(b) # scale values to be positive >> count, bins, ignored = plt.hist(b, 100, normed=True, align=’mid’) >> sigma = np.std(np.log(b)) >> mu = np.mean(np.log(b))

>> x = np.linspace(min(bins), max(bins), 10000) >> pdf = (np.exp(-(np.log(x) - mu)**2 / (2 * sigma**2)) .. / (x * sigma * np.sqrt(2 * np.pi)))

>> plt.plot(x, pdf, color=’r’, linewidth=2) >> plt.show()

dask.array.random.logseries(p, size=None)

Draw samples from a logarithmic series distribution.

Samples are drawn from a log series distribution with specified shape parameter, 0 < p < 1.

Parameters:

p : float or array_like of floats

Shape parameter for the distribution. Must be in the range (0, 1).

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if p is a scalar. Otherwise, np.array(p).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized logarithmic series distribution.

See also

scipy.stats.logser
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Log Series distribution is

\[P(k) = \frac{-p^k}{k \ln(1-p)},\]

where p = probability.

The log series distribution is frequently used to represent species richness and occurrence, first proposed by Fisher, Corbet, and Williams in 1943 [2]. It may also be used to model the numbers of occupants seen in cars [3].

References

[R301304]Buzas, Martin A.; Culver, Stephen J., Understanding regional species diversity through the log series distribution of occurrences: BIODIVERSITY RESEARCH Diversity & Distributions, Volume 5, Number 5, September 1999 , pp. 187-195(9).
[R302304]Fisher, R.A,, A.S. Corbet, and C.B. Williams. 1943. The relation between the number of species and the number of individuals in a random sample of an animal population. Journal of Animal Ecology, 12:42-58.
[R303304]D. J. Hand, F. Daly, D. Lunn, E. Ostrowski, A Handbook of Small Data Sets, CRC Press, 1994.
[R304304]Wikipedia, “Logarithmic distribution”, http://en.wikipedia.org/wiki/Logarithmic_distribution

Examples

Draw samples from the distribution:

>> a = .6 >> s = np.random.logseries(a, 10000) >> count, bins, ignored = plt.hist(s)

# plot against distribution

>> def logseries(k, p): .. return -p**k/(k*log(1-p)) >> plt.plot(bins, logseries(bins, a)*count.max()/

logseries(bins, a).max(), ‘r’)

>> plt.show()

dask.array.random.negative_binomial(n, p, size=None)

Draw samples from a negative binomial distribution.

Samples are drawn from a negative binomial distribution with specified parameters, n trials and p probability of success where n is an integer > 0 and p is in the interval [0, 1].

Parameters:

n : int or array_like of ints

Parameter of the distribution, > 0. Floats are also accepted, but they will be truncated to integers.

p : float or array_like of floats

Parameter of the distribution, >= 0 and <=1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if n and p are both scalars. Otherwise, np.broadcast(n, p).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized negative binomial distribution, where each sample is equal to N, the number of trials it took to achieve n - 1 successes, N - (n - 1) failures, and a success on the, (N + n)th trial.

Notes

The probability density for the negative binomial distribution is

\[P(N;n,p) = \binom{N+n-1}{n-1}p^{n}(1-p)^{N},\]

where \(n-1\) is the number of successes, \(p\) is the probability of success, and \(N+n-1\) is the number of trials. The negative binomial distribution gives the probability of n-1 successes and N failures in N+n-1 trials, and success on the (N+n)th trial.

If one throws a die repeatedly until the third time a “1” appears, then the probability distribution of the number of non-“1”s that appear before the third “1” is a negative binomial distribution.

References

[R309310]Weisstein, Eric W. “Negative Binomial Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/NegativeBinomialDistribution.html
[R310310]Wikipedia, “Negative binomial distribution”, http://en.wikipedia.org/wiki/Negative_binomial_distribution

Examples

Draw samples from the distribution:

A real world example. A company drills wild-cat oil exploration wells, each with an estimated probability of success of 0.1. What is the probability of having one success for each successive well, that is what is the probability of a single success after drilling 5 wells, after 6 wells, etc.?

>> s = np.random.negative_binomial(1, 0.1, 100000) >> for i in range(1, 11): .. probability = sum(s<i) / 100000. .. print i, “wells drilled, probability of one success =”, probability

dask.array.random.noncentral_chisquare(df, nonc, size=None)

Draw samples from a noncentral chi-square distribution.

The noncentral \(\chi^2\) distribution is a generalisation of the \(\chi^2\) distribution.

Parameters:

df : int or array_like of ints

Degrees of freedom, should be > 0 as of NumPy 1.10.0, should be > 1 for earlier versions.

nonc : float or array_like of floats

Non-centrality, should be non-negative.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if df and nonc are both scalars. Otherwise, np.broadcast(df, nonc).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized noncentral chi-square distribution.

Notes

The probability density function for the noncentral Chi-square distribution is

\[P(x;df,nonc) = \sum^{\infty}_{i=0} \frac{e^{-nonc/2}(nonc/2)^{i}}{i!} \P_{Y_{df+2i}}(x),\]

where \(Y_{q}\) is the Chi-square with q degrees of freedom.

In Delhi (2007), it is noted that the noncentral chi-square is useful in bombing and coverage problems, the probability of killing the point target given by the noncentral chi-squared distribution.

References

[R313314]Delhi, M.S. Holla, “On a noncentral chi-square distribution in the analysis of weapon systems effectiveness”, Metrika, Volume 15, Number 1 / December, 1970.
[R314314]Wikipedia, “Noncentral chi-square distribution” http://en.wikipedia.org/wiki/Noncentral_chi-square_distribution

Examples

Draw values from the distribution and plot the histogram

>> import matplotlib.pyplot as plt >> values = plt.hist(np.random.noncentral_chisquare(3, 20, 100000), .. bins=200, normed=True) >> plt.show()

Draw values from a noncentral chisquare with very small noncentrality, and compare to a chisquare.

>> plt.figure() >> values = plt.hist(np.random.noncentral_chisquare(3, .0000001, 100000), .. bins=np.arange(0., 25, .1), normed=True) >> values2 = plt.hist(np.random.chisquare(3, 100000), .. bins=np.arange(0., 25, .1), normed=True) >> plt.plot(values[1][0:-1], values[0]-values2[0], ‘ob’) >> plt.show()

Demonstrate how large values of non-centrality lead to a more symmetric distribution.

>> plt.figure() >> values = plt.hist(np.random.noncentral_chisquare(3, 20, 100000), .. bins=200, normed=True) >> plt.show()

dask.array.random.noncentral_f(dfnum, dfden, nonc, size=None)

Draw samples from the noncentral F distribution.

Samples are drawn from an F distribution with specified parameters, dfnum (degrees of freedom in numerator) and dfden (degrees of freedom in denominator), where both parameters > 1. nonc is the non-centrality parameter.

Parameters:

dfnum : int or array_like of ints

Parameter, should be > 1.

dfden : int or array_like of ints

Parameter, should be > 1.

nonc : float or array_like of floats

Parameter, should be >= 0.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if dfnum, dfden, and nonc are all scalars. Otherwise, np.broadcast(dfnum, dfden, nonc).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized noncentral Fisher distribution.

Notes

When calculating the power of an experiment (power = probability of rejecting the null hypothesis when a specific alternative is true) the non-central F statistic becomes important. When the null hypothesis is true, the F statistic follows a central F distribution. When the null hypothesis is not true, then it follows a non-central F statistic.

References

[R317318]Weisstein, Eric W. “Noncentral F-Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/NoncentralF-Distribution.html
[R318318]Wikipedia, “Noncentral F-distribution”, http://en.wikipedia.org/wiki/Noncentral_F-distribution

Examples

In a study, testing for a specific alternative to the null hypothesis requires use of the Noncentral F distribution. We need to calculate the area in the tail of the distribution that exceeds the value of the F distribution for the null hypothesis. We’ll plot the two probability distributions for comparison.

>> dfnum = 3 # between group deg of freedom >> dfden = 20 # within groups degrees of freedom >> nonc = 3.0 >> nc_vals = np.random.noncentral_f(dfnum, dfden, nonc, 1000000) >> NF = np.histogram(nc_vals, bins=50, normed=True) >> c_vals = np.random.f(dfnum, dfden, 1000000) >> F = np.histogram(c_vals, bins=50, normed=True) >> plt.plot(F[1][1:], F[0]) >> plt.plot(NF[1][1:], NF[0]) >> plt.show()

dask.array.random.normal(loc=0.0, scale=1.0, size=None)

Draw random samples from a normal (Gaussian) distribution.

The probability density function of the normal distribution, first derived by De Moivre and 200 years later by both Gauss and Laplace independently [R322322], is often called the bell curve because of its characteristic shape (see the example below).

The normal distributions occurs often in nature. For example, it describes the commonly occurring distribution of samples influenced by a large number of tiny, random disturbances, each with its own unique distribution [R322322].

Parameters:

loc : float or array_like of floats

Mean (“centre”) of the distribution.

scale : float or array_like of floats

Standard deviation (spread or “width”) of the distribution.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if loc and scale are both scalars. Otherwise, np.broadcast(loc, scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized normal distribution.

See also

scipy.stats.norm
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Gaussian distribution is

\[p(x) = \frac{1}{\sqrt{ 2 \pi \sigma^2 }} e^{ - \frac{ (x - \mu)^2 } {2 \sigma^2} },\]

where \(\mu\) is the mean and \(\sigma\) the standard deviation. The square of the standard deviation, \(\sigma^2\), is called the variance.

The function has its peak at the mean, and its “spread” increases with the standard deviation (the function reaches 0.607 times its maximum at \(x + \sigma\) and \(x - \sigma\) [R322322]). This implies that numpy.random.normal is more likely to return samples lying close to the mean, rather than those far away.

References

[R321322]Wikipedia, “Normal distribution”, http://en.wikipedia.org/wiki/Normal_distribution
[R322322](1, 2, 3, 4) P. R. Peebles Jr., “Central Limit Theorem” in “Probability, Random Variables and Random Signal Principles”, 4th ed., 2001, pp. 51, 51, 125.

Examples

Draw samples from the distribution:

>> mu, sigma = 0, 0.1 # mean and standard deviation >> s = np.random.normal(mu, sigma, 1000)

Verify the mean and the variance:

>> abs(mu - np.mean(s)) < 0.01 True

>> abs(sigma - np.std(s, ddof=1)) < 0.01 True

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, 30, normed=True) >> plt.plot(bins, 1/(sigma * np.sqrt(2 * np.pi)) * .. np.exp( - (bins - mu)**2 / (2 * sigma**2) ), .. linewidth=2, color=’r’) >> plt.show()

dask.array.random.pareto(a, size=None)

Draw samples from a Pareto II or Lomax distribution with specified shape.

The Lomax or Pareto II distribution is a shifted Pareto distribution. The classical Pareto distribution can be obtained from the Lomax distribution by adding 1 and multiplying by the scale parameter m (see Notes). The smallest value of the Lomax distribution is zero while for the classical Pareto distribution it is mu, where the standard Pareto distribution has location mu = 1. Lomax can also be considered as a simplified version of the Generalized Pareto distribution (available in SciPy), with the scale set to one and the location set to zero.

The Pareto distribution must be greater than zero, and is unbounded above. It is also known as the “80-20 rule”. In this distribution, 80 percent of the weights are in the lowest 20 percent of the range, while the other 20 percent fill the remaining 80 percent of the range.

Parameters:

a : float or array_like of floats

Shape of the distribution. Should be greater than zero.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if a is a scalar. Otherwise, np.array(a).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Pareto distribution.

See also

scipy.stats.lomax
probability density function, distribution or cumulative density function, etc.
scipy.stats.genpareto
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Pareto distribution is

\[p(x) = \frac{am^a}{x^{a+1}}\]

where \(a\) is the shape and \(m\) the scale.

The Pareto distribution, named after the Italian economist Vilfredo Pareto, is a power law probability distribution useful in many real world problems. Outside the field of economics it is generally referred to as the Bradford distribution. Pareto developed the distribution to describe the distribution of wealth in an economy. It has also found use in insurance, web page access statistics, oil field sizes, and many other problems, including the download frequency for projects in Sourceforge [R325328]. It is one of the so-called “fat-tailed” distributions.

References

[R325328](1, 2) Francis Hunt and Paul Johnson, On the Pareto Distribution of Sourceforge projects.
[R326328]Pareto, V. (1896). Course of Political Economy. Lausanne.
[R327328]Reiss, R.D., Thomas, M.(2001), Statistical Analysis of Extreme Values, Birkhauser Verlag, Basel, pp 23-30.
[R328328]Wikipedia, “Pareto distribution”, http://en.wikipedia.org/wiki/Pareto_distribution

Examples

Draw samples from the distribution:

>> a, m = 3., 2. # shape and mode >> s = (np.random.pareto(a, 1000) + 1) * m

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, _ = plt.hist(s, 100, normed=True) >> fit = a*m**a / bins**(a+1) >> plt.plot(bins, max(count)*fit/max(fit), linewidth=2, color=’r’) >> plt.show()

dask.array.random.poisson(lam=1.0, size=None)

Draw samples from a Poisson distribution.

The Poisson distribution is the limit of the binomial distribution for large N.

Parameters:

lam : float or array_like of floats

Expectation of interval, should be >= 0. A sequence of expectation intervals must be broadcastable over the requested size.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if lam is a scalar. Otherwise, np.array(lam).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Poisson distribution.

Notes

The Poisson distribution

\[f(k; \lambda)=\frac{\lambda^k e^{-\lambda}}{k!}\]

For events with an expected separation \(\lambda\) the Poisson distribution \(f(k; \lambda)\) describes the probability of \(k\) events occurring within the observed interval \(\lambda\).

Because the output is limited to the range of the C long type, a ValueError is raised when lam is within 10 sigma of the maximum representable value.

References

[R333334]Weisstein, Eric W. “Poisson Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/PoissonDistribution.html
[R334334]Wikipedia, “Poisson distribution”, http://en.wikipedia.org/wiki/Poisson_distribution

Examples

Draw samples from the distribution:

>> import numpy as np >> s = np.random.poisson(5, 10000)

Display histogram of the sample:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, 14, normed=True) >> plt.show()

Draw each 100 values for lambda 100 and 500:

>> s = np.random.poisson(lam=(100., 500.), size=(100, 2))

dask.array.random.power(a, size=None)

Draws samples in [0, 1] from a power distribution with positive exponent a - 1.

Also known as the power function distribution.

Parameters:

a : float or array_like of floats

Parameter of the distribution. Should be greater than zero.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if a is a scalar. Otherwise, np.array(a).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized power distribution.
Raises:

ValueError

If a < 1.

Notes

The probability density function is

\[P(x; a) = ax^{a-1}, 0 \le x \le 1, a>0.\]

The power function distribution is just the inverse of the Pareto distribution. It may also be seen as a special case of the Beta distribution.

It is used, for example, in modeling the over-reporting of insurance claims.

References

[R337338]Christian Kleiber, Samuel Kotz, “Statistical size distributions in economics and actuarial sciences”, Wiley, 2003.
[R338338]Heckert, N. A. and Filliben, James J. “NIST Handbook 148: Dataplot Reference Manual, Volume 2: Let Subcommands and Library Functions”, National Institute of Standards and Technology Handbook Series, June 2003. http://www.itl.nist.gov/div898/software/dataplot/refman2/auxillar/powpdf.pdf

Examples

Draw samples from the distribution:

>> a = 5. # shape >> samples = 1000 >> s = np.random.power(a, samples)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, bins=30) >> x = np.linspace(0, 1, 100) >> y = a*x**(a-1.) >> normed_y = samples*np.diff(bins)[0]*y >> plt.plot(x, normed_y) >> plt.show()

Compare the power function distribution to the inverse of the Pareto.

>> from scipy import stats >> rvs = np.random.power(5, 1000000) >> rvsp = np.random.pareto(5, 1000000) >> xx = np.linspace(0,1,100) >> powpdf = stats.powerlaw.pdf(xx,5)

>> plt.figure() >> plt.hist(rvs, bins=50, normed=True) >> plt.plot(xx,powpdf,’r-‘) >> plt.title(‘np.random.power(5)’)

>> plt.figure() >> plt.hist(1./(1.+rvsp), bins=50, normed=True) >> plt.plot(xx,powpdf,’r-‘) >> plt.title(‘inverse of 1 + np.random.pareto(5)’)

>> plt.figure() >> plt.hist(1./(1.+rvsp), bins=50, normed=True) >> plt.plot(xx,powpdf,’r-‘) >> plt.title(‘inverse of stats.pareto(5)’)

dask.array.random.random(size=None)

Return random floats in the half-open interval [0.0, 1.0).

Results are from the “continuous uniform” distribution over the stated interval. To sample \(Unif[a, b), b > a\) multiply the output of random_sample by (b-a) and add a:

(b - a) * random_sample() + a
Parameters:

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. Default is None, in which case a single value is returned.
Returns:

out : float or ndarray of floats

Array of random floats of shape size (unless size=None, in which case a single float is returned).

Examples

>> np.random.random_sample() 0.47108547995356098 >> type(np.random.random_sample()) <type ‘float’> >> np.random.random_sample((5,)) array([ 0.30220482, 0.86820401, 0.1654503 , 0.11659149, 0.54323428])

Three-by-two array of random numbers from [-5, 0):

>> 5 * np.random.random_sample((3, 2)) - 5 array([[-3.99149989, -0.52338984],

[-2.99091858, -0.79479508], [-1.23204345, -1.75224494]])
dask.array.random.random_sample(size=None)

Return random floats in the half-open interval [0.0, 1.0).

Results are from the “continuous uniform” distribution over the stated interval. To sample \(Unif[a, b), b > a\) multiply the output of random_sample by (b-a) and add a:

(b - a) * random_sample() + a
Parameters:

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. Default is None, in which case a single value is returned.
Returns:

out : float or ndarray of floats

Array of random floats of shape size (unless size=None, in which case a single float is returned).

Examples

>> np.random.random_sample() 0.47108547995356098 >> type(np.random.random_sample()) <type ‘float’> >> np.random.random_sample((5,)) array([ 0.30220482, 0.86820401, 0.1654503 , 0.11659149, 0.54323428])

Three-by-two array of random numbers from [-5, 0):

>> 5 * np.random.random_sample((3, 2)) - 5 array([[-3.99149989, -0.52338984],

[-2.99091858, -0.79479508], [-1.23204345, -1.75224494]])
dask.array.random.rayleigh(scale=1.0, size=None)

Draw samples from a Rayleigh distribution.

The \(\chi\) and Weibull distributions are generalizations of the Rayleigh.

Parameters:

scale : float or array_like of floats, optional

Scale, also equals the mode. Should be >= 0. Default is 1.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if scale is a scalar. Otherwise, np.array(scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Rayleigh distribution.

Notes

The probability density function for the Rayleigh distribution is

\[P(x;scale) = \frac{x}{scale^2}e^{\frac{-x^2}{2 \cdotp scale^2}}\]

The Rayleigh distribution would arise, for example, if the East and North components of the wind velocity had identical zero-mean Gaussian distributions. Then the wind speed would have a Rayleigh distribution.

References

[R341342]Brighton Webs Ltd., “Rayleigh Distribution,” http://www.brighton-webs.co.uk/distributions/rayleigh.asp
[R342342]Wikipedia, “Rayleigh distribution” http://en.wikipedia.org/wiki/Rayleigh_distribution

Examples

Draw values from the distribution and plot the histogram

>> values = hist(np.random.rayleigh(3, 100000), bins=200, normed=True)

Wave heights tend to follow a Rayleigh distribution. If the mean wave height is 1 meter, what fraction of waves are likely to be larger than 3 meters?

>> meanvalue = 1 >> modevalue = np.sqrt(2 / np.pi) * meanvalue >> s = np.random.rayleigh(modevalue, 1000000)

The percentage of waves larger than 3 meters is:

>> 100.*sum(s>3)/1000000. 0.087300000000000003

dask.array.random.standard_cauchy(size=None)

Draw samples from a standard Cauchy distribution with mode = 0.

Also known as the Lorentz distribution.

Parameters:

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. Default is None, in which case a single value is returned.
Returns:

samples : ndarray or scalar

The drawn samples.

Notes

The probability density function for the full Cauchy distribution is

\[P(x; x_0, \gamma) = \frac{1}{\pi \gamma \bigl[ 1+ (\frac{x-x_0}{\gamma})^2 \bigr] }\]

and the Standard Cauchy distribution just sets \(x_0=0\) and \(\gamma=1\)

The Cauchy distribution arises in the solution to the driven harmonic oscillator problem, and also describes spectral line broadening. It also describes the distribution of values at which a line tilted at a random angle will cut the x axis.

When studying hypothesis tests that assume normality, seeing how the tests perform on data from a Cauchy distribution is a good indicator of their sensitivity to a heavy-tailed distribution, since the Cauchy looks very much like a Gaussian distribution, but with heavier tails.

References

[R345347]NIST/SEMATECH e-Handbook of Statistical Methods, “Cauchy Distribution”, http://www.itl.nist.gov/div898/handbook/eda/section3/eda3663.htm
[R346347]Weisstein, Eric W. “Cauchy Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/CauchyDistribution.html
[R347347]Wikipedia, “Cauchy distribution” http://en.wikipedia.org/wiki/Cauchy_distribution

Examples

Draw samples and plot the distribution:

>> s = np.random.standard_cauchy(1000000) >> s = s[(s>-25) & (s<25)] # truncate distribution so it plots well >> plt.hist(s, bins=100) >> plt.show()

dask.array.random.standard_exponential(size=None)

Draw samples from the standard exponential distribution.

standard_exponential is identical to the exponential distribution with a scale parameter of 1.

Parameters:

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. Default is None, in which case a single value is returned.
Returns:

out : float or ndarray

Drawn samples.

Examples

Output a 3x8000 array:

>> n = np.random.standard_exponential((3, 8000))

dask.array.random.standard_gamma(shape, size=None)

Draw samples from a standard Gamma distribution.

Samples are drawn from a Gamma distribution with specified parameters, shape (sometimes designated “k”) and scale=1.

Parameters:

shape : float or array_like of floats

Parameter, should be > 0.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if shape is a scalar. Otherwise, np.array(shape).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized standard gamma distribution.

See also

scipy.stats.gamma
probability density function, distribution or cumulative density function, etc.

Notes

The probability density for the Gamma distribution is

\[p(x) = x^{k-1}\frac{e^{-x/\theta}}{\theta^k\Gamma(k)},\]

where \(k\) is the shape and \(\theta\) the scale, and \(\Gamma\) is the Gamma function.

The Gamma distribution is often used to model the times to failure of electronic components, and arises naturally in processes for which the waiting times between Poisson distributed events are relevant.

References

[R351352]Weisstein, Eric W. “Gamma Distribution.” From MathWorld–A Wolfram Web Resource. http://mathworld.wolfram.com/GammaDistribution.html
[R352352]Wikipedia, “Gamma distribution”, http://en.wikipedia.org/wiki/Gamma_distribution

Examples

Draw samples from the distribution:

>> shape, scale = 2., 1. # mean and width >> s = np.random.standard_gamma(shape, 1000000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> import scipy.special as sps >> count, bins, ignored = plt.hist(s, 50, normed=True) >> y = bins**(shape-1) * ((np.exp(-bins/scale))/ .. (sps.gamma(shape) * scale**shape)) >> plt.plot(bins, y, linewidth=2, color=’r’) >> plt.show()

dask.array.random.standard_normal(size=None)

Draw samples from a standard Normal distribution (mean=0, stdev=1).

Parameters:

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. Default is None, in which case a single value is returned.
Returns:

out : float or ndarray

Drawn samples.

Examples

>> s = np.random.standard_normal(8000) >> s array([ 0.6888893 , 0.78096262, -0.89086505, .., 0.49876311, #random

-0.38672696, -0.4685006 ]) #random

>> s.shape (8000,) >> s = np.random.standard_normal(size=(3, 4, 2)) >> s.shape (3, 4, 2)

dask.array.random.standard_t(df, size=None)

Draw samples from a standard Student’s t distribution with df degrees of freedom.

A special case of the hyperbolic distribution. As df gets large, the result resembles that of the standard normal distribution (standard_normal).

Parameters:

df : int or array_like of ints

Degrees of freedom, should be > 0.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if df is a scalar. Otherwise, np.array(df).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized standard Student’s t distribution.

Notes

The probability density function for the t distribution is

\[P(x, df) = \frac{\Gamma(\frac{df+1}{2})}{\sqrt{\pi df} \Gamma(\frac{df}{2})}\Bigl( 1+\frac{x^2}{df} \Bigr)^{-(df+1)/2}\]

The t test is based on an assumption that the data come from a Normal distribution. The t test provides a way to test whether the sample mean (that is the mean calculated from the data) is a good estimate of the true mean.

The derivation of the t-distribution was first published in 1908 by William Gisset while working for the Guinness Brewery in Dublin. Due to proprietary issues, he had to publish under a pseudonym, and so he used the name Student.

References

[R355356](1, 2) Dalgaard, Peter, “Introductory Statistics With R”, Springer, 2002.
[R356356]Wikipedia, “Student’s t-distribution” http://en.wikipedia.org/wiki/Student’s_t-distribution

Examples

From Dalgaard page 83 [R355356], suppose the daily energy intake for 11 women in Kj is:

>> intake = np.array([5260., 5470, 5640, 6180, 6390, 6515, 6805, 7515, .. 7515, 8230, 8770])

Does their energy intake deviate systematically from the recommended value of 7725 kJ?

We have 10 degrees of freedom, so is the sample mean within 95% of the recommended value?

>> s = np.random.standard_t(10, size=100000) >> np.mean(intake) 6753.636363636364 >> intake.std(ddof=1) 1142.1232221373727

Calculate the t statistic, setting the ddof parameter to the unbiased value so the divisor in the standard deviation will be degrees of freedom, N-1.

>> t = (np.mean(intake)-7725)/(intake.std(ddof=1)/np.sqrt(len(intake))) >> import matplotlib.pyplot as plt >> h = plt.hist(s, bins=100, normed=True)

For a one-sided t-test, how far out in the distribution does the t statistic appear?

>> np.sum(s<t) / float(len(s)) 0.0090699999999999999 #random

So the p-value is about 0.009, which says the null hypothesis has a probability of about 99% of being true.

dask.array.random.triangular(left, mode, right, size=None)

Draw samples from the triangular distribution over the interval [left, right].

The triangular distribution is a continuous probability distribution with lower limit left, peak at mode, and upper limit right. Unlike the other distributions, these parameters directly define the shape of the pdf.

Parameters:

left : float or array_like of floats

Lower limit.

mode : float or array_like of floats

The value where the peak of the distribution occurs. The value should fulfill the condition left <= mode <= right.

right : float or array_like of floats

Upper limit, should be larger than left.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if left, mode, and right are all scalars. Otherwise, np.broadcast(left, mode, right).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized triangular distribution.

Notes

The probability density function for the triangular distribution is

\[\begin{split}P(x;l, m, r) = \begin{cases} \frac{2(x-l)}{(r-l)(m-l)}& \text{for $l \leq x \leq m$},\\ \frac{2(r-x)}{(r-l)(r-m)}& \text{for $m \leq x \leq r$},\\ 0& \text{otherwise}. \end{cases}\end{split}\]

The triangular distribution is often used in ill-defined problems where the underlying distribution is not known, but some knowledge of the limits and mode exists. Often it is used in simulations.

References

[R359359]Wikipedia, “Triangular distribution” http://en.wikipedia.org/wiki/Triangular_distribution

Examples

Draw values from the distribution and plot the histogram:

>> import matplotlib.pyplot as plt >> h = plt.hist(np.random.triangular(-3, 0, 8, 100000), bins=200, .. normed=True) >> plt.show()

dask.array.random.uniform(low=0.0, high=1.0, size=None)

Draw samples from a uniform distribution.

Samples are uniformly distributed over the half-open interval [low, high) (includes low, but excludes high). In other words, any value within the given interval is equally likely to be drawn by uniform.

Parameters:

low : float or array_like of floats, optional

Lower boundary of the output interval. All values generated will be greater than or equal to low. The default value is 0.

high : float or array_like of floats

Upper boundary of the output interval. All values generated will be less than high. The default value is 1.0.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if low and high are both scalars. Otherwise, np.broadcast(low, high).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized uniform distribution.

See also

randint
Discrete uniform distribution, yielding integers.
random_integers
Discrete uniform distribution over the closed interval [low, high].
random_sample
Floats uniformly distributed over [0, 1).
random
Alias for random_sample.
rand
Convenience function that accepts dimensions as input, e.g., rand(2,2) would generate a 2-by-2 array of floats, uniformly distributed over [0, 1).

Notes

The probability density function of the uniform distribution is

\[p(x) = \frac{1}{b - a}\]

anywhere within the interval [a, b), and zero elsewhere.

When high == low, values of low will be returned. If high < low, the results are officially undefined and may eventually raise an error, i.e. do not rely on this function to behave when passed arguments satisfying that inequality condition.

Examples

Draw samples from the distribution:

>> s = np.random.uniform(-1,0,1000)

All values are within the given interval:

>> np.all(s >= -1) True >> np.all(s < 0) True

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> count, bins, ignored = plt.hist(s, 15, normed=True) >> plt.plot(bins, np.ones_like(bins), linewidth=2, color=’r’) >> plt.show()

dask.array.random.vonmises(mu, kappa, size=None)

Draw samples from a von Mises distribution.

Samples are drawn from a von Mises distribution with specified mode (mu) and dispersion (kappa), on the interval [-pi, pi].

The von Mises distribution (also known as the circular normal distribution) is a continuous probability distribution on the unit circle. It may be thought of as the circular analogue of the normal distribution.

Parameters:

mu : float or array_like of floats

Mode (“center”) of the distribution.

kappa : float or array_like of floats

Dispersion of the distribution, has to be >=0.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if mu and kappa are both scalars. Otherwise, np.broadcast(mu, kappa).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized von Mises distribution.

See also

scipy.stats.vonmises
probability density function, distribution, or cumulative density function, etc.

Notes

The probability density for the von Mises distribution is

\[p(x) = \frac{e^{\kappa cos(x-\mu)}}{2\pi I_0(\kappa)},\]

where \(\mu\) is the mode and \(\kappa\) the dispersion, and \(I_0(\kappa)\) is the modified Bessel function of order 0.

The von Mises is named for Richard Edler von Mises, who was born in Austria-Hungary, in what is now the Ukraine. He fled to the United States in 1939 and became a professor at Harvard. He worked in probability theory, aerodynamics, fluid mechanics, and philosophy of science.

References

[R361362]Abramowitz, M. and Stegun, I. A. (Eds.). “Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, 9th printing,” New York: Dover, 1972.
[R362362]von Mises, R., “Mathematical Theory of Probability and Statistics”, New York: Academic Press, 1964.

Examples

Draw samples from the distribution:

>> mu, kappa = 0.0, 4.0 # mean and dispersion >> s = np.random.vonmises(mu, kappa, 1000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> from scipy.special import i0 >> plt.hist(s, 50, normed=True) >> x = np.linspace(-np.pi, np.pi, num=51) >> y = np.exp(kappa*np.cos(x-mu))/(2*np.pi*i0(kappa)) >> plt.plot(x, y, linewidth=2, color=’r’) >> plt.show()

dask.array.random.wald(mean, scale, size=None)

Draw samples from a Wald, or inverse Gaussian, distribution.

As the scale approaches infinity, the distribution becomes more like a Gaussian. Some references claim that the Wald is an inverse Gaussian with mean equal to 1, but this is by no means universal.

The inverse Gaussian distribution was first studied in relationship to Brownian motion. In 1956 M.C.K. Tweedie used the name inverse Gaussian because there is an inverse relationship between the time to cover a unit distance and distance covered in unit time.

Parameters:

mean : float or array_like of floats

Distribution mean, should be > 0.

scale : float or array_like of floats

Scale parameter, should be >= 0.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if mean and scale are both scalars. Otherwise, np.broadcast(mean, scale).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Wald distribution.

Notes

The probability density function for the Wald distribution is

\[P(x;mean,scale) = \sqrt{\frac{scale}{2\pi x^3}}e^ \frac{-scale(x-mean)^2}{2\cdotp mean^2x}\]

As noted above the inverse Gaussian distribution first arise from attempts to model Brownian motion. It is also a competitor to the Weibull for use in reliability modeling and modeling stock returns and interest rate processes.

References

[R365367]Brighton Webs Ltd., Wald Distribution, http://www.brighton-webs.co.uk/distributions/wald.asp
[R366367]Chhikara, Raj S., and Folks, J. Leroy, “The Inverse Gaussian Distribution: Theory : Methodology, and Applications”, CRC Press, 1988.
[R367367]Wikipedia, “Wald distribution” http://en.wikipedia.org/wiki/Wald_distribution

Examples

Draw values from the distribution and plot the histogram:

>> import matplotlib.pyplot as plt >> h = plt.hist(np.random.wald(3, 2, 100000), bins=200, normed=True) >> plt.show()

dask.array.random.weibull(a, size=None)

Draw samples from a Weibull distribution.

Draw samples from a 1-parameter Weibull distribution with the given shape parameter a.

\[X = (-ln(U))^{1/a}\]

Here, U is drawn from the uniform distribution over (0,1].

The more common 2-parameter Weibull, including a scale parameter \(\lambda\) is just \(X = \lambda(-ln(U))^{1/a}\).

Parameters:

a : float or array_like of floats

Shape of the distribution. Should be greater than zero.

size : int or tuple of ints, optional

Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn. If size is None (default), a single value is returned if a is a scalar. Otherwise, np.array(a).size samples are drawn.
Returns:

out : ndarray or scalar

Drawn samples from the parameterized Weibull distribution.

See also

scipy.stats.weibull_max, scipy.stats.weibull_min, scipy.stats.genextreme, gumbel

Notes

The Weibull (or Type III asymptotic extreme value distribution for smallest values, SEV Type III, or Rosin-Rammler distribution) is one of a class of Generalized Extreme Value (GEV) distributions used in modeling extreme value problems. This class includes the Gumbel and Frechet distributions.

The probability density for the Weibull distribution is

\[p(x) = \frac{a} {\lambda}(\frac{x}{\lambda})^{a-1}e^{-(x/\lambda)^a},\]

where \(a\) is the shape and \(\lambda\) the scale.

The function has its peak (the mode) at \(\lambda(\frac{a-1}{a})^{1/a}\).

When a = 1, the Weibull distribution reduces to the exponential distribution.

References

[R371373]Waloddi Weibull, Royal Technical University, Stockholm, 1939 “A Statistical Theory Of The Strength Of Materials”, Ingeniorsvetenskapsakademiens Handlingar Nr 151, 1939, Generalstabens Litografiska Anstalts Forlag, Stockholm.
[R372373]Waloddi Weibull, “A Statistical Distribution Function of Wide Applicability”, Journal Of Applied Mechanics ASME Paper 1951.
[R373373]Wikipedia, “Weibull distribution”, http://en.wikipedia.org/wiki/Weibull_distribution

Examples

Draw samples from the distribution:

>> a = 5. # shape >> s = np.random.weibull(a, 1000)

Display the histogram of the samples, along with the probability density function:

>> import matplotlib.pyplot as plt >> x = np.arange(1,100.)/50. >> def weib(x,n,a): .. return (a / n) * (x / n)**(a - 1) * np.exp(-(x / n)**a)

>> count, bins, ignored = plt.hist(np.random.weibull(5.,1000)) >> x = np.arange(1,100.)/50. >> scale = count.max()/weib(x, 1., 5.).max() >> plt.plot(x, weib(x, 1., 5.)*scale) >> plt.show()

dask.array.random.zipf(a, size=None)

Standard distributions

dask.array.stats.ttest_ind(a, b, axis=0, equal_var=True)

Calculates the T-test for the means of two independent samples of scores.

This is a two-sided test for the null hypothesis that 2 independent samples have identical average (expected) values. This test assumes that the populations have identical variances by default.

Parameters:

a, b : array_like

The arrays must have the same shape, except in the dimension corresponding to axis (the first, by default).

axis : int or None, optional

Axis along which to compute test. If None, compute over the whole arrays, a, and b.

equal_var : bool, optional

If True (default), perform a standard independent 2 sample test that assumes equal population variances [R377378]. If False, perform Welch’s t-test, which does not assume equal population variance [R378378].

New in version 0.11.0.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

statistic : float or array

The calculated t-statistic.

pvalue : float or array

The two-tailed p-value.

Notes

We can use this test, if we observe two independent samples from the same or different population, e.g. exam scores of boys and girls or of two ethnic groups. The test measures whether the average (expected) value differs significantly across samples. If we observe a large p-value, for example larger than 0.05 or 0.1, then we cannot reject the null hypothesis of identical average scores. If the p-value is smaller than the threshold, e.g. 1%, 5% or 10%, then we reject the null hypothesis of equal averages.

References

[R377378](1, 2) http://en.wikipedia.org/wiki/T-test#Independent_two-sample_t-test
[R378378](1, 2) http://en.wikipedia.org/wiki/Welch%27s_t_test

Examples

>> from scipy import stats >> np.random.seed(12345678)

Test with sample with identical means:

>> rvs1 = stats.norm.rvs(loc=5,scale=10,size=500) >> rvs2 = stats.norm.rvs(loc=5,scale=10,size=500) >> stats.ttest_ind(rvs1,rvs2) (0.26833823296239279, 0.78849443369564776) >> stats.ttest_ind(rvs1,rvs2, equal_var = False) (0.26833823296239279, 0.78849452749500748)

ttest_ind underestimates p for unequal variances:

>> rvs3 = stats.norm.rvs(loc=5, scale=20, size=500) >> stats.ttest_ind(rvs1, rvs3) (-0.46580283298287162, 0.64145827413436174) >> stats.ttest_ind(rvs1, rvs3, equal_var = False) (-0.46580283298287162, 0.64149646246569292)

When n1 != n2, the equal variance t-statistic is no longer equal to the unequal variance t-statistic:

>> rvs4 = stats.norm.rvs(loc=5, scale=20, size=100) >> stats.ttest_ind(rvs1, rvs4) (-0.99882539442782481, 0.3182832709103896) >> stats.ttest_ind(rvs1, rvs4, equal_var = False) (-0.69712570584654099, 0.48716927725402048)

T-test with different means, variance, and n:

>> rvs5 = stats.norm.rvs(loc=8, scale=20, size=100) >> stats.ttest_ind(rvs1, rvs5) (-1.4679669854490653, 0.14263895620529152) >> stats.ttest_ind(rvs1, rvs5, equal_var = False) (-0.94365973617132992, 0.34744170334794122)

dask.array.stats.ttest_1samp(a, popmean, axis=0, nan_policy='propagate')

Calculates the T-test for the mean of ONE group of scores.

This is a two-sided test for the null hypothesis that the expected value (mean) of a sample of independent observations a is equal to the given population mean, popmean.

Parameters:

a : array_like

sample observation

popmean : float or array_like

expected value in null hypothesis, if array_like than it must have the same shape as a excluding the axis dimension

axis : int or None, optional

Axis along which to compute test. If None, compute over the whole array a.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

statistic : float or array

t-statistic

pvalue : float or array

two-tailed p-value

Examples

>> from scipy import stats

>> np.random.seed(7654567) # fix seed to get the same result >> rvs = stats.norm.rvs(loc=5, scale=10, size=(50,2))

Test if mean of random sample is equal to true mean, and different mean. We reject the null hypothesis in the second case and don’t reject it in the first case.

>> stats.ttest_1samp(rvs,5.0) (array([-0.68014479, -0.04323899]), array([ 0.49961383, 0.96568674])) >> stats.ttest_1samp(rvs,0.0) (array([ 2.77025808, 4.11038784]), array([ 0.00789095, 0.00014999]))

Examples using axis and non-scalar dimension for population mean.

>> stats.ttest_1samp(rvs,[5.0,0.0]) (array([-0.68014479, 4.11038784]), array([ 4.99613833e-01, 1.49986458e-04])) >> stats.ttest_1samp(rvs.T,[5.0,0.0],axis=1) (array([-0.68014479, 4.11038784]), array([ 4.99613833e-01, 1.49986458e-04])) >> stats.ttest_1samp(rvs,[[5.0],[0.0]]) (array([[-0.68014479, -0.04323899],

[ 2.77025808, 4.11038784]]), array([[ 4.99613833e-01, 9.65686743e-01], [ 7.89094663e-03, 1.49986458e-04]]))
dask.array.stats.ttest_rel(a, b, axis=0, nan_policy='propagate')

Calculates the T-test on TWO RELATED samples of scores, a and b.

This is a two-sided test for the null hypothesis that 2 related or repeated samples have identical average (expected) values.

Parameters:

a, b : array_like

The arrays must have the same shape.

axis : int or None, optional

Axis along which to compute test. If None, compute over the whole arrays, a, and b.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

statistic : float or array

t-statistic

pvalue : float or array

two-tailed p-value

Notes

Examples for the use are scores of the same set of student in different exams, or repeated sampling from the same units. The test measures whether the average score differs significantly across samples (e.g. exams). If we observe a large p-value, for example greater than 0.05 or 0.1 then we cannot reject the null hypothesis of identical average scores. If the p-value is smaller than the threshold, e.g. 1%, 5% or 10%, then we reject the null hypothesis of equal averages. Small p-values are associated with large t-statistics.

References

http://en.wikipedia.org/wiki/T-test#Dependent_t-test

Examples

>> from scipy import stats >> np.random.seed(12345678) # fix random seed to get same numbers

>> rvs1 = stats.norm.rvs(loc=5,scale=10,size=500) >> rvs2 = (stats.norm.rvs(loc=5,scale=10,size=500) + .. stats.norm.rvs(scale=0.2,size=500)) >> stats.ttest_rel(rvs1,rvs2) (0.24101764965300962, 0.80964043445811562) >> rvs3 = (stats.norm.rvs(loc=8,scale=10,size=500) + .. stats.norm.rvs(scale=0.2,size=500)) >> stats.ttest_rel(rvs1,rvs3) (-3.9995108708727933, 7.3082402191726459e-005)

dask.array.stats.chisquare(f_obs, f_exp=None, ddof=0, axis=0)

Calculates a one-way chi square test.

The chi square test tests the null hypothesis that the categorical data has the given frequencies.

Parameters:

f_obs : array_like

Observed frequencies in each category.

f_exp : array_like, optional

Expected frequencies in each category. By default the categories are assumed to be equally likely.

ddof : int, optional

“Delta degrees of freedom”: adjustment to the degrees of freedom for the p-value. The p-value is computed using a chi-squared distribution with k - 1 - ddof degrees of freedom, where k is the number of observed frequencies. The default value of ddof is 0.

axis : int or None, optional

The axis of the broadcast result of f_obs and f_exp along which to apply the test. If axis is None, all values in f_obs are treated as a single data set. Default is 0.
Returns:

chisq : float or ndarray

The chi-squared test statistic. The value is a float if axis is None or f_obs and f_exp are 1-D.

p : float or ndarray

The p-value of the test. The value is a float if ddof and the return value chisq are scalars.

See also

power_divergence, mstats.chisquare

Notes

This test is invalid when the observed or expected frequencies in each category are too small. A typical rule is that all of the observed and expected frequencies should be at least 5.

The default degrees of freedom, k-1, are for the case when no parameters of the distribution are estimated. If p parameters are estimated by efficient maximum likelihood then the correct degrees of freedom are k-1-p. If the parameters are estimated in a different way, then the dof can be between k-1-p and k-1. However, it is also possible that the asymptotic distribution is not a chisquare, in which case this test is not appropriate.

References

[R381382]Lowry, Richard. “Concepts and Applications of Inferential Statistics”. Chapter 8. http://faculty.vassar.edu/lowry/ch8pt1.html
[R382382]“Chi-squared test”, http://en.wikipedia.org/wiki/Chi-squared_test

Examples

When just f_obs is given, it is assumed that the expected frequencies are uniform and given by the mean of the observed frequencies.

>> from scipy.stats import chisquare >> chisquare([16, 18, 16, 14, 12, 12]) (2.0, 0.84914503608460956)

With f_exp the expected frequencies can be given.

>> chisquare([16, 18, 16, 14, 12, 12], f_exp=[16, 16, 16, 16, 16, 8]) (3.5, 0.62338762774958223)

When f_obs is 2-D, by default the test is applied to each column.

>> obs = np.array([[16, 18, 16, 14, 12, 12], [32, 24, 16, 28, 20, 24]]).T >> obs.shape (6, 2) >> chisquare(obs) (array([ 2. , 6.66666667]), array([ 0.84914504, 0.24663415]))

By setting axis=None, the test is applied to all data in the array, which is equivalent to applying the test to the flattened array.

>> chisquare(obs, axis=None) (23.31034482758621, 0.015975692534127565) >> chisquare(obs.ravel()) (23.31034482758621, 0.015975692534127565)

ddof is the change to make to the default degrees of freedom.

>> chisquare([16, 18, 16, 14, 12, 12], ddof=1) (2.0, 0.73575888234288467)

The calculation of the p-values is done by broadcasting the chi-squared statistic with ddof.

>> chisquare([16, 18, 16, 14, 12, 12], ddof=[0,1,2]) (2.0, array([ 0.84914504, 0.73575888, 0.5724067 ]))

f_obs and f_exp are also broadcast. In the following, f_obs has shape (6,) and f_exp has shape (2, 6), so the result of broadcasting f_obs and f_exp has shape (2, 6). To compute the desired chi-squared statistics, we use axis=1:

>> chisquare([16, 18, 16, 14, 12, 12], .. f_exp=[[16, 16, 16, 16, 16, 8], [8, 20, 20, 16, 12, 12]], .. axis=1) (array([ 3.5 , 9.25]), array([ 0.62338763, 0.09949846]))

dask.array.stats.power_divergence(f_obs, f_exp=None, ddof=0, axis=0, lambda_=None)

Cressie-Read power divergence statistic and goodness of fit test.

This function tests the null hypothesis that the categorical data has the given frequencies, using the Cressie-Read power divergence statistic.

Parameters:

f_obs : array_like

Observed frequencies in each category.

f_exp : array_like, optional

Expected frequencies in each category. By default the categories are assumed to be equally likely.

ddof : int, optional

“Delta degrees of freedom”: adjustment to the degrees of freedom for the p-value. The p-value is computed using a chi-squared distribution with k - 1 - ddof degrees of freedom, where k is the number of observed frequencies. The default value of ddof is 0.

axis : int or None, optional

The axis of the broadcast result of f_obs and f_exp along which to apply the test. If axis is None, all values in f_obs are treated as a single data set. Default is 0.

lambda_ : float or str, optional

lambda_ gives the power in the Cressie-Read power divergence statistic. The default is 1. For convenience, lambda_ may be assigned one of the following strings, in which case the corresponding numerical value is used:

String              Value   Description
"pearson"             1     Pearson's chi-squared statistic.
                            In this case, the function is
                            equivalent to `stats.chisquare`.
"log-likelihood"      0     Log-likelihood ratio. Also known as
                            the G-test [R387389]_.
"freeman-tukey"      -1/2   Freeman-Tukey statistic.
"mod-log-likelihood" -1     Modified log-likelihood ratio.
"neyman"             -2     Neyman's statistic.
"cressie-read"        2/3   The power recommended in [R389389]_.
Returns:

statistic : float or ndarray

The Cressie-Read power divergence test statistic. The value is a float if axis is None or if` f_obs and f_exp are 1-D.

pvalue : float or ndarray

The p-value of the test. The value is a float if ddof and the return value stat are scalars.

See also

chisquare

Notes

This test is invalid when the observed or expected frequencies in each category are too small. A typical rule is that all of the observed and expected frequencies should be at least 5.

When lambda_ is less than zero, the formula for the statistic involves dividing by f_obs, so a warning or error may be generated if any value in f_obs is 0.

Similarly, a warning or error may be generated if any value in f_exp is zero when lambda_ >= 0.

The default degrees of freedom, k-1, are for the case when no parameters of the distribution are estimated. If p parameters are estimated by efficient maximum likelihood then the correct degrees of freedom are k-1-p. If the parameters are estimated in a different way, then the dof can be between k-1-p and k-1. However, it is also possible that the asymptotic distribution is not a chisquare, in which case this test is not appropriate.

This function handles masked arrays. If an element of f_obs or f_exp is masked, then data at that position is ignored, and does not count towards the size of the data set.

New in version 0.13.0.

References

[R385389]Lowry, Richard. “Concepts and Applications of Inferential Statistics”. Chapter 8. http://faculty.vassar.edu/lowry/ch8pt1.html
[R386389]“Chi-squared test”, http://en.wikipedia.org/wiki/Chi-squared_test
[R387389]“G-test”, http://en.wikipedia.org/wiki/G-test
[R388389]Sokal, R. R. and Rohlf, F. J. “Biometry: the principles and practice of statistics in biological research”, New York: Freeman (1981)
[R389389]Cressie, N. and Read, T. R. C., “Multinomial Goodness-of-Fit Tests”, J. Royal Stat. Soc. Series B, Vol. 46, No. 3 (1984), pp. 440-464.

Examples

(See chisquare for more examples.)

When just f_obs is given, it is assumed that the expected frequencies are uniform and given by the mean of the observed frequencies. Here we perform a G-test (i.e. use the log-likelihood ratio statistic):

>> from scipy.stats import power_divergence >> power_divergence([16, 18, 16, 14, 12, 12], lambda_=’log-likelihood’) (2.006573162632538, 0.84823476779463769)

The expected frequencies can be given with the f_exp argument:

>> power_divergence([16, 18, 16, 14, 12, 12], .. f_exp=[16, 16, 16, 16, 16, 8], .. lambda_=’log-likelihood’) (3.3281031458963746, 0.6495419288047497)

When f_obs is 2-D, by default the test is applied to each column.

>> obs = np.array([[16, 18, 16, 14, 12, 12], [32, 24, 16, 28, 20, 24]]).T >> obs.shape (6, 2) >> power_divergence(obs, lambda_=”log-likelihood”) (array([ 2.00657316, 6.77634498]), array([ 0.84823477, 0.23781225]))

By setting axis=None, the test is applied to all data in the array, which is equivalent to applying the test to the flattened array.

>> power_divergence(obs, axis=None) (23.31034482758621, 0.015975692534127565) >> power_divergence(obs.ravel()) (23.31034482758621, 0.015975692534127565)

ddof is the change to make to the default degrees of freedom.

>> power_divergence([16, 18, 16, 14, 12, 12], ddof=1) (2.0, 0.73575888234288467)

The calculation of the p-values is done by broadcasting the test statistic with ddof.

>> power_divergence([16, 18, 16, 14, 12, 12], ddof=[0,1,2]) (2.0, array([ 0.84914504, 0.73575888, 0.5724067 ]))

f_obs and f_exp are also broadcast. In the following, f_obs has shape (6,) and f_exp has shape (2, 6), so the result of broadcasting f_obs and f_exp has shape (2, 6). To compute the desired chi-squared statistics, we must use axis=1:

>> power_divergence([16, 18, 16, 14, 12, 12], .. f_exp=[[16, 16, 16, 16, 16, 8], .. [8, 20, 20, 16, 12, 12]], .. axis=1) (array([ 3.5 , 9.25]), array([ 0.62338763, 0.09949846]))

dask.array.stats.skew(a, axis=0, bias=True, nan_policy='propagate')

Computes the skewness of a data set.

For normally distributed data, the skewness should be about 0. A skewness value > 0 means that there is more weight in the left tail of the distribution. The function skewtest can be used to determine if the skewness value is close enough to 0, statistically speaking.

Parameters:

a : ndarray

data

axis : int or None, optional

Axis along which skewness is calculated. Default is 0. If None, compute over the whole array a.

bias : bool, optional

If False, then the calculations are corrected for statistical bias.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

skewness : ndarray

The skewness of values along an axis, returning 0 where all values are equal.

References

[R395395]Zwillinger, D. and Kokoska, S. (2000). CRC Standard Probability and Statistics Tables and Formulae. Chapman & Hall: New York. 2000. Section 2.2.24.1
dask.array.stats.skewtest(a, axis=0, nan_policy='propagate')

Tests whether the skew is different from the normal distribution.

This function tests the null hypothesis that the skewness of the population that the sample was drawn from is the same as that of a corresponding normal distribution.

Parameters:

a : array

The data to be tested

axis : int or None, optional

Axis along which statistics are calculated. Default is 0. If None, compute over the whole array a.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

statistic : float

The computed z-score for this test.

pvalue : float

a 2-sided p-value for the hypothesis test

Notes

The sample size must be at least 8.

dask.array.stats.kurtosis(a, axis=0, fisher=True, bias=True, nan_policy='propagate')

Computes the kurtosis (Fisher or Pearson) of a dataset.

Kurtosis is the fourth central moment divided by the square of the variance. If Fisher’s definition is used, then 3.0 is subtracted from the result to give 0.0 for a normal distribution.

If bias is False then the kurtosis is calculated using k statistics to eliminate bias coming from biased moment estimators

Use kurtosistest to see if result is close enough to normal.

Parameters:

a : array

data for which the kurtosis is calculated

axis : int or None, optional

Axis along which the kurtosis is calculated. Default is 0. If None, compute over the whole array a.

fisher : bool, optional

If True, Fisher’s definition is used (normal ==> 0.0). If False, Pearson’s definition is used (normal ==> 3.0).

bias : bool, optional

If False, then the calculations are corrected for statistical bias.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

kurtosis : array

The kurtosis of values along an axis. If all values are equal, return -3 for Fisher’s definition and 0 for Pearson’s definition.

References

[R397397]Zwillinger, D. and Kokoska, S. (2000). CRC Standard Probability and Statistics Tables and Formulae. Chapman & Hall: New York. 2000.
dask.array.stats.kurtosistest(a, axis=0, nan_policy='propagate')

Tests whether a dataset has normal kurtosis

This function tests the null hypothesis that the kurtosis of the population from which the sample was drawn is that of the normal distribution: kurtosis = 3(n-1)/(n+1).

Parameters:

a : array

array of the sample data

axis : int or None, optional

Axis along which to compute test. Default is 0. If None, compute over the whole array a.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

statistic : float

The computed z-score for this test.

pvalue : float

The 2-sided p-value for the hypothesis test

Notes

Valid only for n>20. The Z-score is set to 0 for bad entries. This function uses the method described in [R399399].

References

[R399399](1, 2) see e.g. F. J. Anscombe, W. J. Glynn, “Distribution of the kurtosis statistic b2 for normal samples”, Biometrika, vol. 70, pp. 227-234, 1983.
dask.array.stats.normaltest(a, axis=0, nan_policy='propagate')

Tests whether a sample differs from a normal distribution.

This function tests the null hypothesis that a sample comes from a normal distribution. It is based on D’Agostino and Pearson’s [R401402], [R402402] test that combines skew and kurtosis to produce an omnibus test of normality.

Parameters:

a : array_like

The array containing the data to be tested.

axis : int or None, optional

Axis along which to compute test. Default is 0. If None, compute over the whole array a.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

statistic : float or array

s^2 + k^2, where s is the z-score returned by skewtest and k is the z-score returned by kurtosistest.

pvalue : float or array

A 2-sided chi squared probability for the hypothesis test.

References

[R401402](1, 2) D’Agostino, R. B. (1971), “An omnibus test of normality for moderate and large sample size”, Biometrika, 58, 341-348
[R402402](1, 2) D’Agostino, R. and Pearson, E. S. (1973), “Tests for departure from normality”, Biometrika, 60, 613-622
dask.array.stats.f_oneway(*args)

Performs a 1-way ANOVA.

The one-way ANOVA tests the null hypothesis that two or more groups have the same population mean. The test is applied to samples from two or more groups, possibly with differing sizes.

Parameters:

sample1, sample2, .. : array_like

The sample measurements for each group.
Returns:

statistic : float

The computed F-value of the test.

pvalue : float

The associated p-value from the F-distribution.

Notes

The ANOVA test has important assumptions that must be satisfied in order for the associated p-value to be valid.

  1. The samples are independent.
  2. Each sample is from a normally distributed population.
  3. The population standard deviations of the groups are all equal. This property is known as homoscedasticity.

If these assumptions are not true for a given set of data, it may still be possible to use the Kruskal-Wallis H-test (scipy.stats.kruskal) although with some loss of power.

The algorithm is from Heiman[2], pp.394-7.

References

[R405407]Lowry, Richard. “Concepts and Applications of Inferential Statistics”. Chapter 14. http://faculty.vassar.edu/lowry/ch14pt1.html
[R406407]Heiman, G.W. Research Methods in Statistics. 2002.
[R407407](1, 2) McDonald, G. H. “Handbook of Biological Statistics”, One-way ANOVA. http://www.biostathandbook.com/onewayanova.html

Examples

>> import scipy.stats as stats

[R407407] Here are some data on a shell measurement (the length of the anterior adductor muscle scar, standardized by dividing by length) in the mussel Mytilus trossulus from five locations: Tillamook, Oregon; Newport, Oregon; Petersburg, Alaska; Magadan, Russia; and Tvarminne, Finland, taken from a much larger data set used in McDonald et al. (1991).

>> tillamook = [0.0571, 0.0813, 0.0831, 0.0976, 0.0817, 0.0859, 0.0735, .. 0.0659, 0.0923, 0.0836] >> newport = [0.0873, 0.0662, 0.0672, 0.0819, 0.0749, 0.0649, 0.0835, .. 0.0725] >> petersburg = [0.0974, 0.1352, 0.0817, 0.1016, 0.0968, 0.1064, 0.105] >> magadan = [0.1033, 0.0915, 0.0781, 0.0685, 0.0677, 0.0697, 0.0764, .. 0.0689] >> tvarminne = [0.0703, 0.1026, 0.0956, 0.0973, 0.1039, 0.1045] >> stats.f_oneway(tillamook, newport, petersburg, magadan, tvarminne) (7.1210194716424473, 0.00028122423145345439)

dask.array.stats.moment(a, moment=1, axis=0, nan_policy='propagate')

Calculates the nth moment about the mean for a sample.

A moment is a specific quantitative measure of the shape of a set of points. It is often used to calculate coefficients of skewness and kurtosis due to its close relationship with them.

Parameters:

a : array_like

data

moment : int or array_like of ints, optional

order of central moment that is returned. Default is 1.

axis : int or None, optional

Axis along which the central moment is computed. Default is 0. If None, compute over the whole array a.

nan_policy : {‘propagate’, ‘raise’, ‘omit’}, optional

Defines how to handle when input contains nan. ‘propagate’ returns nan, ‘raise’ throws an error, ‘omit’ performs the calculations ignoring nan values. Default is ‘propagate’.
Returns:

n-th central moment : ndarray or float

The appropriate moment along the given axis or over all values if axis is None. The denominator for the moment calculation is the number of observations, no degrees of freedom correction is done.

See also

kurtosis, skew, describe

Notes

The k-th central moment of a data sample is:

\[m_k = \frac{1}{n} \sum_{i = 1}^n (x_i - \bar{x})^k\]

Where n is the number of samples and x-bar is the mean. This function uses exponentiation by squares [R411411] for efficiency.

References

[R411411](1, 2) http://eli.thegreenplace.net/2009/03/21/efficient-integer-exponentiation-algorithms
dask.array.image.imread(filename, imread=None, preprocess=None)

Read a stack of images into a dask array

Parameters:

filename: string

A globstring like ‘myfile.*.png’

imread: function (optional)

Optionally provide custom imread function. Function should expect a filename and produce a numpy array. Defaults to skimage.io.imread.

preprocess: function (optional)

Optionally provide custom function to preprocess the image. Function should expect a numpy array for a single image.
Returns:

Dask array of all images stacked along the first dimension. All images

will be treated as individual chunks

Examples

>>> from dask.array.image import imread
>>> im = imread('2015-*-*.png')  
>>> im.shape  
(365, 1000, 1000, 3)
dask.array.core.map_blocks(func, *args, **kwargs)

Map a function across all blocks of a dask array.

Parameters:

func : callable

Function to apply to every block in the array.

args : dask arrays or constants

dtype : np.dtype, optional

The dtype of the output array. It is recommended to provide this. If not provided, will be inferred by applying the function to a small set of fake data.

chunks : tuple, optional

Chunk shape of resulting blocks if the function does not preserve shape. If not provided, the resulting array is assumed to have the same block structure as the first input array.

drop_axis : number or iterable, optional

Dimensions lost by the function.

new_axis : number or iterable, optional

New dimensions created by the function. Note that these are applied after drop_axis (if present).

token : string, optional

The key prefix to use for the output array. If not provided, will be determined from the function name.

name : string, optional

The key name to use for the output array. Note that this fully specifies the output key name, and must be unique. If not provided, will be determined by a hash of the arguments.

**kwargs :

Other keyword arguments to pass to function. Values must be constants (not dask.arrays)

Examples

>>> import dask.array as da
>>> x = da.arange(6, chunks=3)

>>> x.map_blocks(lambda x: x * 2).compute()
array([ 0,  2,  4,  6,  8, 10])

The da.map_blocks function can also accept multiple arrays.

>>> d = da.arange(5, chunks=2)
>>> e = da.arange(5, chunks=2)

>>> f = map_blocks(lambda a, b: a + b**2, d, e)
>>> f.compute()
array([ 0,  2,  6, 12, 20])

If the function changes shape of the blocks then you must provide chunks explicitly.

>>> y = x.map_blocks(lambda x: x[::2], chunks=((2, 2),))

You have a bit of freedom in specifying chunks. If all of the output chunk sizes are the same, you can provide just that chunk size as a single tuple.

>>> a = da.arange(18, chunks=(6,))
>>> b = a.map_blocks(lambda x: x[:3], chunks=(3,))

If the function changes the dimension of the blocks you must specify the created or destroyed dimensions.

>>> b = a.map_blocks(lambda x: x[None, :, None], chunks=(1, 6, 1),
...                  new_axis=[0, 2])

Map_blocks aligns blocks by block positions without regard to shape. In the following example we have two arrays with the same number of blocks but with different shape and chunk sizes.

>>> x = da.arange(1000, chunks=(100,))
>>> y = da.arange(100, chunks=(10,))

The relevant attribute to match is numblocks.

>>> x.numblocks
(10,)
>>> y.numblocks
(10,)

If these match (up to broadcasting rules) then we can map arbitrary functions across blocks

>>> def func(a, b):
...     return np.array([a.max(), b.max()])

>>> da.map_blocks(func, x, y, chunks=(2,), dtype='i8')
dask.array<func, shape=(20,), dtype=int64, chunksize=(2,)>

>>> _.compute()
array([ 99,   9, 199,  19, 299,  29, 399,  39, 499,  49, 599,  59, 699,
        69, 799,  79, 899,  89, 999,  99])

Your block function can learn where in the array it is if it supports a block_id keyword argument. This will receive entries like (2, 0, 1), the position of the block in the dask array.

>>> def func(block, block_id=None):
...     pass

You may specify the key name prefix of the resulting task in the graph with the optional token keyword argument.

>>> x.map_blocks(lambda x: x + 1, token='increment')  
dask.array<increment, shape=(100,), dtype=int64, chunksize=(10,)>
dask.array.core.atop(func, out_ind, *args, **kwargs)

Tensor operation: Generalized inner and outer products

A broad class of blocked algorithms and patterns can be specified with a concise multi-index notation. The atop function applies an in-memory function across multiple blocks of multiple inputs in a variety of ways. Many dask.array operations are special cases of atop including elementwise, broadcasting, reductions, tensordot, and transpose.

Parameters:

func : callable

Function to apply to individual tuples of blocks

out_ind : iterable

Block pattern of the output, something like ‘ijk’ or (1, 2, 3)

*args : sequence of Array, index pairs

Sequence like (x, ‘ij’, y, ‘jk’, z, ‘i’)

**kwargs : dict

Extra keyword arguments to pass to function

dtype : np.dtype

Datatype of resulting array.

concatenate : bool, keyword only

If true concatenate arrays along dummy indices, else provide lists

adjust_chunks : dict

Dictionary mapping index to function to be applied to chunk sizes

new_axes : dict, keyword only

New indexes and their dimension lengths

See also

top, contains

Examples

2D embarrassingly parallel operation from two arrays, x, and y.

>>> z = atop(operator.add, 'ij', x, 'ij', y, 'ij', dtype='f8')  # z = x + y

Outer product multiplying x by y, two 1-d vectors

>>> z = atop(operator.mul, 'ij', x, 'i', y, 'j', dtype='f8')

z = x.T

>>> z = atop(np.transpose, 'ji', x, 'ij', dtype=x.dtype)

The transpose case above is illustrative because it does same transposition both on each in-memory block by calling np.transpose and on the order of the blocks themselves, by switching the order of the index ij -> ji.

We can compose these same patterns with more variables and more complex in-memory functions

z = X + Y.T

>>> z = atop(lambda x, y: x + y.T, 'ij', x, 'ij', y, 'ji', dtype='f8')

Any index, like i missing from the output index is interpreted as a contraction (note that this differs from Einstein convention; repeated indices do not imply contraction.) In the case of a contraction the passed function should expect an iterable of blocks on any array that holds that index. To receive arrays concatenated along contracted dimensions instead pass concatenate=True.

Inner product multiplying x by y, two 1-d vectors

>>> def sequence_dot(x_blocks, y_blocks):
...     result = 0
...     for x, y in zip(x_blocks, y_blocks):
...         result += x.dot(y)
...     return result

>>> z = atop(sequence_dot, '', x, 'i', y, 'i', dtype='f8')

Add new single-chunk dimensions with the new_axes= keyword, including the length of the new dimension. New dimensions will always be in a single chunk.

>>> def f(x):
...     return x[:, None] * np.ones((1, 5))

>>> z = atop(f, 'az', x, 'a', new_axes={'z': 5}, dtype=x.dtype)

If the applied function changes the size of each chunk you can specify this with a adjust_chunks={...} dictionary holding a function for each index that modifies the dimension size in that index.

>>> def double(x):
...     return np.concatenate([x, x])

>>> y = atop(double, 'ij', x, 'ij',
...          adjust_chunks={'i': lambda n: 2 * n}, dtype=x.dtype)
dask.array.core.top(func, output, out_indices, *arrind_pairs, **kwargs)

Tensor operation

Applies a function, func, across blocks from many different input dasks. We arrange the pattern with which those blocks interact with sets of matching indices. E.g.:

top(func, 'z', 'i', 'x', 'i', 'y', 'i')

yield an embarrassingly parallel communication pattern and is read as

$$ z_i = func(x_i, y_i) $$

More complex patterns may emerge, including multiple indices:

top(func, 'z', 'ij', 'x', 'ij', 'y', 'ji')

$$ z_{ij} = func(x_{ij}, y_{ji}) $$

Indices missing in the output but present in the inputs results in many inputs being sent to one function (see examples).

See also

atop

Examples

Simple embarrassing map operation

>>> inc = lambda x: x + 1
>>> top(inc, 'z', 'ij', 'x', 'ij', numblocks={'x': (2, 2)})  
{('z', 0, 0): (inc, ('x', 0, 0)),
 ('z', 0, 1): (inc, ('x', 0, 1)),
 ('z', 1, 0): (inc, ('x', 1, 0)),
 ('z', 1, 1): (inc, ('x', 1, 1))}

Simple operation on two datasets

>>> add = lambda x, y: x + y
>>> top(add, 'z', 'ij', 'x', 'ij', 'y', 'ij', numblocks={'x': (2, 2),
...                                                      'y': (2, 2)})  
{('z', 0, 0): (add, ('x', 0, 0), ('y', 0, 0)),
 ('z', 0, 1): (add, ('x', 0, 1), ('y', 0, 1)),
 ('z', 1, 0): (add, ('x', 1, 0), ('y', 1, 0)),
 ('z', 1, 1): (add, ('x', 1, 1), ('y', 1, 1))}

Operation that flips one of the datasets

>>> addT = lambda x, y: x + y.T  # Transpose each chunk
>>> #                                        z_ij ~ x_ij y_ji
>>> #               ..         ..         .. notice swap
>>> top(addT, 'z', 'ij', 'x', 'ij', 'y', 'ji', numblocks={'x': (2, 2),
...                                                       'y': (2, 2)})  
{('z', 0, 0): (add, ('x', 0, 0), ('y', 0, 0)),
 ('z', 0, 1): (add, ('x', 0, 1), ('y', 1, 0)),
 ('z', 1, 0): (add, ('x', 1, 0), ('y', 0, 1)),
 ('z', 1, 1): (add, ('x', 1, 1), ('y', 1, 1))}

Dot product with contraction over j index. Yields list arguments

>>> top(dotmany, 'z', 'ik', 'x', 'ij', 'y', 'jk', numblocks={'x': (2, 2),
...                                                          'y': (2, 2)})  
{('z', 0, 0): (dotmany, [('x', 0, 0), ('x', 0, 1)],
                        [('y', 0, 0), ('y', 1, 0)]),
 ('z', 0, 1): (dotmany, [('x', 0, 0), ('x', 0, 1)],
                        [('y', 0, 1), ('y', 1, 1)]),
 ('z', 1, 0): (dotmany, [('x', 1, 0), ('x', 1, 1)],
                        [('y', 0, 0), ('y', 1, 0)]),
 ('z', 1, 1): (dotmany, [('x', 1, 0), ('x', 1, 1)],
                        [('y', 0, 1), ('y', 1, 1)])}

Pass concatenate=True to concatenate arrays ahead of time

>>> top(f, 'z', 'i', 'x', 'ij', 'y', 'ij', concatenate=True,
...     numblocks={'x': (2, 2), 'y': (2, 2,)})  
{('z', 0): (f, (concatenate_axes, [('x', 0, 0), ('x', 0, 1)], (1,)),
               (concatenate_axes, [('y', 0, 0), ('y', 0, 1)], (1,)))
 ('z', 1): (f, (concatenate_axes, [('x', 1, 0), ('x', 1, 1)], (1,)),
               (concatenate_axes, [('y', 1, 0), ('y', 1, 1)], (1,)))}

Supports Broadcasting rules

>>> top(add, 'z', 'ij', 'x', 'ij', 'y', 'ij', numblocks={'x': (1, 2),
...                                                      'y': (2, 2)})  
{('z', 0, 0): (add, ('x', 0, 0), ('y', 0, 0)),
 ('z', 0, 1): (add, ('x', 0, 1), ('y', 0, 1)),
 ('z', 1, 0): (add, ('x', 0, 0), ('y', 1, 0)),
 ('z', 1, 1): (add, ('x', 0, 1), ('y', 1, 1))}

Support keyword arguments with apply

>>> def f(a, b=0): return a + b
>>> top(f, 'z', 'i', 'x', 'i', numblocks={'x': (2,)}, b=10)  
{('z', 0): (apply, f, [('x', 0)], {'b': 10}),
 ('z', 1): (apply, f, [('x', 1)], {'b': 10})}

Array Methods

class dask.array.Array

Parallel Dask Array

A parallel nd-array comprised of many numpy arrays arranged in a grid.

This constructor is for advanced uses only. For normal use see the da.from_array function.

Parameters:

dask : dict

Task dependency graph

name : string

Name of array in dask

shape : tuple of ints

Shape of the entire array

chunks: iterable of tuples

block sizes along each dimension

See also

dask.array.from_array

all(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Test whether all array elements along a given axis evaluate to True.

Parameters:

a : array_like

Input array or object that can be converted to an array.

axis : None or int or tuple of ints, optional

Axis or axes along which a logical AND reduction is performed. The default (axis = None) is to perform a logical AND over all the dimensions of the input array. axis may be negative, in which case it counts from the last to the first axis.

New in version 1.7.0.

If this is a tuple of ints, a reduction is performed on multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output and its type is preserved (e.g., if dtype(out) is float, the result will consist of 0.0’s and 1.0’s). See doc.ufuncs (Section “Output arguments”) for more details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the all method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

all : ndarray, bool

A new boolean or array is returned unless out is specified, in which case a reference to out is returned.

See also

ndarray.all
equivalent method
any
Test whether any element along a given axis evaluates to True.

Notes

Not a Number (NaN), positive infinity and negative infinity evaluate to True because these are not equal to zero.

Examples

>>> np.all([[True,False],[True,True]])
False

>>> np.all([[True,False],[True,True]], axis=0)
array([ True, False], dtype=bool)

>>> np.all([-1, 4, 5])
True

>>> np.all([1.0, np.nan])
True

>>> o=np.array([False])
>>> z=np.all([-1, 4, 5], out=o)
>>> id(z), id(o), z                             
(28293632, 28293632, array([ True], dtype=bool))
any(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Test whether any array element along a given axis evaluates to True.

Returns single boolean unless axis is not None

Parameters:

a : array_like

Input array or object that can be converted to an array.

axis : None or int or tuple of ints, optional

Axis or axes along which a logical OR reduction is performed. The default (axis = None) is to perform a logical OR over all the dimensions of the input array. axis may be negative, in which case it counts from the last to the first axis.

New in version 1.7.0.

If this is a tuple of ints, a reduction is performed on multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output and its type is preserved (e.g., if it is of type float, then it will remain so, returning 1.0 for True and 0.0 for False, regardless of the type of a). See doc.ufuncs (Section “Output arguments”) for details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the any method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

any : bool or ndarray

A new boolean or ndarray is returned unless out is specified, in which case a reference to out is returned.

See also

ndarray.any
equivalent method
all
Test whether all elements along a given axis evaluate to True.

Notes

Not a Number (NaN), positive infinity and negative infinity evaluate to True because these are not equal to zero.

Examples

>>> np.any([[True, False], [True, True]])
True

>>> np.any([[True, False], [False, False]], axis=0)
array([ True, False], dtype=bool)

>>> np.any([-1, 0, 5])
True

>>> np.any(np.nan)
True

>>> o=np.array([False])
>>> z=np.any([-1, 4, 5], out=o)
>>> z, o
(array([ True], dtype=bool), array([ True], dtype=bool))
>>> # Check now that z is a reference to o
>>> z is o
True
>>> id(z), id(o) # identity of z and o              
(191614240, 191614240)
argmax(a, axis=None, out=None)

Returns the indices of the maximum values along an axis.

Parameters:

a : array_like

Input array.

axis : int, optional

By default, the index is into the flattened array, otherwise along the specified axis.

out : array, optional

If provided, the result will be inserted into this array. It should be of the appropriate shape and dtype.
Returns:

index_array : ndarray of ints

Array of indices into the array. It has the same shape as a.shape with the dimension along axis removed.

See also

ndarray.argmax, argmin

amax
The maximum value along a given axis.
unravel_index
Convert a flat index into an index tuple.

Notes

In case of multiple occurrences of the maximum values, the indices corresponding to the first occurrence are returned.

Examples

>>> a = np.arange(6).reshape(2,3)
>>> a
array([[0, 1, 2],
       [3, 4, 5]])
>>> np.argmax(a)
5
>>> np.argmax(a, axis=0)
array([1, 1, 1])
>>> np.argmax(a, axis=1)
array([2, 2])

>>> b = np.arange(6)
>>> b[1] = 5
>>> b
array([0, 5, 2, 3, 4, 5])
>>> np.argmax(b) # Only the first occurrence is returned.
1
argmin(a, axis=None, out=None)

Returns the indices of the minimum values along an axis.

Parameters:

a : array_like

Input array.

axis : int, optional

By default, the index is into the flattened array, otherwise along the specified axis.

out : array, optional

If provided, the result will be inserted into this array. It should be of the appropriate shape and dtype.
Returns:

index_array : ndarray of ints

Array of indices into the array. It has the same shape as a.shape with the dimension along axis removed.

See also

ndarray.argmin, argmax

amin
The minimum value along a given axis.
unravel_index
Convert a flat index into an index tuple.

Notes

In case of multiple occurrences of the minimum values, the indices corresponding to the first occurrence are returned.

Examples

>>> a = np.arange(6).reshape(2,3)
>>> a
array([[0, 1, 2],
       [3, 4, 5]])
>>> np.argmin(a)
0
>>> np.argmin(a, axis=0)
array([0, 0, 0])
>>> np.argmin(a, axis=1)
array([0, 0])

>>> b = np.arange(6)
>>> b[4] = 0
>>> b
array([0, 1, 2, 3, 0, 5])
>>> np.argmin(b) # Only the first occurrence is returned.
0
astype(dtype, **kwargs)

Copy of the array, cast to a specified type.

Parameters:

dtype : str or dtype

Typecode or data-type to which the array is cast.

casting : {‘no’, ‘equiv’, ‘safe’, ‘same_kind’, ‘unsafe’}, optional

Controls what kind of data casting may occur. Defaults to ‘unsafe’ for backwards compatibility.

  • ‘no’ means the data types should not be cast at all.
  • ‘equiv’ means only byte-order changes are allowed.
  • ‘safe’ means only casts which can preserve values are allowed.
  • ‘same_kind’ means only safe casts or casts within a kind,
    like float64 to float32, are allowed.
  • ‘unsafe’ means any data conversions may be done.

copy : bool, optional

By default, astype always returns a newly allocated array. If this is set to False and the dtype requirement is satisfied, the input array is returned instead of a copy.
clip(a, a_min, a_max, out=None)

Clip (limit) the values in an array.

Given an interval, values outside the interval are clipped to the interval edges. For example, if an interval of [0, 1] is specified, values smaller than 0 become 0, and values larger than 1 become 1.

Parameters:

a : array_like

Array containing elements to clip.

a_min : scalar or array_like

Minimum value.

a_max : scalar or array_like

Maximum value. If a_min or a_max are array_like, then they will be broadcasted to the shape of a.

out : ndarray, optional

The results will be placed in this array. It may be the input array for in-place clipping. out must be of the right shape to hold the output. Its type is preserved.
Returns:

clipped_array : ndarray

An array with the elements of a, but where values < a_min are replaced with a_min, and those > a_max with a_max.

See also

numpy.doc.ufuncs
Section “Output arguments”

Examples

>>> a = np.arange(10)
>>> np.clip(a, 1, 8)
array([1, 1, 2, 3, 4, 5, 6, 7, 8, 8])
>>> a
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
>>> np.clip(a, 3, 6, out=a)
array([3, 3, 3, 3, 4, 5, 6, 6, 6, 6])
>>> a = np.arange(10)
>>> a
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
>>> np.clip(a, [3,4,1,1,1,4,4,4,4,4], 8)
array([3, 4, 2, 3, 4, 5, 6, 7, 8, 8])
copy()

Copy array. This is a no-op for dask.arrays, which are immutable

cumprod(axis, dtype=None, out=None)

See da.cumprod for docstring

cumsum(axis, dtype=None, out=None)

See da.cumsum for docstring

dot(a, b, out=None)

Dot product of two arrays.

For 2-D arrays it is equivalent to matrix multiplication, and for 1-D arrays to inner product of vectors (without complex conjugation). For N dimensions it is a sum product over the last axis of a and the second-to-last of b:

dot(a, b)[i,j,k,m] = sum(a[i,j,:] * b[k,:,m])
Parameters:

a : array_like

First argument.

b : array_like

Second argument.

out : ndarray, optional

Output argument. This must have the exact kind that would be returned if it was not used. In particular, it must have the right type, must be C-contiguous, and its dtype must be the dtype that would be returned for dot(a,b). This is a performance feature. Therefore, if these conditions are not met, an exception is raised, instead of attempting to be flexible.
Returns:

output : ndarray

Returns the dot product of a and b. If a and b are both scalars or both 1-D arrays then a scalar is returned; otherwise an array is returned. If out is given, then it is returned.
Raises:

ValueError

If the last dimension of a is not the same size as the second-to-last dimension of b.

See also

vdot
Complex-conjugating dot product.
tensordot
Sum products over arbitrary axes.
einsum
Einstein summation convention.
matmul
‘@’ operator as method with out parameter.

Examples

>>> np.dot(3, 4)
12

Neither argument is complex-conjugated:

>>> np.dot([2j, 3j], [2j, 3j])
(-13+0j)

For 2-D arrays it is the matrix product:

>>> a = [[1, 0], [0, 1]]
>>> b = [[4, 1], [2, 2]]
>>> np.dot(a, b)
array([[4, 1],
       [2, 2]])

>>> a = np.arange(3*4*5*6).reshape((3,4,5,6))
>>> b = np.arange(3*4*5*6)[::-1].reshape((5,4,6,3))
>>> np.dot(a, b)[2,3,2,1,2,2]
499128
>>> sum(a[2,3,2,:] * b[1,2,:,2])
499128
flatten(a, order='C')

Return a contiguous flattened array.

A 1-D array, containing the elements of the input, is returned. A copy is made only if needed.

As of NumPy 1.10, the returned array will have the same type as the input array. (for example, a masked array will be returned for a masked array input)

Parameters:

a : array_like

Input array. The elements in a are read in the order specified by order, and packed as a 1-D array.

order : {‘C’,’F’, ‘A’, ‘K’}, optional

The elements of a are read using this index order. ‘C’ means to index the elements in row-major, C-style order, with the last axis index changing fastest, back to the first axis index changing slowest. ‘F’ means to index the elements in column-major, Fortran-style order, with the first index changing fastest, and the last index changing slowest. Note that the ‘C’ and ‘F’ options take no account of the memory layout of the underlying array, and only refer to the order of axis indexing. ‘A’ means to read the elements in Fortran-like index order if a is Fortran contiguous in memory, C-like order otherwise. ‘K’ means to read the elements in the order they occur in memory, except for reversing the data when strides are negative. By default, ‘C’ index order is used.
Returns:

y : array_like

If a is a matrix, y is a 1-D ndarray, otherwise y is an array of the same subtype as a. The shape of the returned array is (a.size,). Matrices are special cased for backward compatibility.

See also

ndarray.flat
1-D iterator over an array.
ndarray.flatten
1-D array copy of the elements of an array in row-major order.
ndarray.reshape
Change the shape of an array without changing its data.

Notes

In row-major, C-style order, in two dimensions, the row index varies the slowest, and the column index the quickest. This can be generalized to multiple dimensions, where row-major order implies that the index along the first axis varies slowest, and the index along the last quickest. The opposite holds for column-major, Fortran-style index ordering.

When a view is desired in as many cases as possible, arr.reshape(-1) may be preferable.

Examples

It is equivalent to reshape(-1, order=order).

>>> x = np.array([[1, 2, 3], [4, 5, 6]])
>>> print(np.ravel(x))
[1 2 3 4 5 6]

>>> print(x.reshape(-1))
[1 2 3 4 5 6]

>>> print(np.ravel(x, order='F'))
[1 4 2 5 3 6]

When order is ‘A’, it will preserve the array’s ‘C’ or ‘F’ ordering:

>>> print(np.ravel(x.T))
[1 4 2 5 3 6]
>>> print(np.ravel(x.T, order='A'))
[1 2 3 4 5 6]

When order is ‘K’, it will preserve orderings that are neither ‘C’ nor ‘F’, but won’t reverse axes:

>>> a = np.arange(3)[::-1]; a
array([2, 1, 0])
>>> a.ravel(order='C')
array([2, 1, 0])
>>> a.ravel(order='K')
array([2, 1, 0])

>>> a = np.arange(12).reshape(2,3,2).swapaxes(1,2); a
array([[[ 0,  2,  4],
        [ 1,  3,  5]],
       [[ 6,  8, 10],
        [ 7,  9, 11]]])
>>> a.ravel(order='C')
array([ 0,  2,  4,  1,  3,  5,  6,  8, 10,  7,  9, 11])
>>> a.ravel(order='K')
array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11])
itemsize

Length of one array element in bytes

map_blocks(func, *args, **kwargs)

Map a function across all blocks of a dask array.

Parameters:

func : callable

Function to apply to every block in the array.

args : dask arrays or constants

dtype : np.dtype, optional

The dtype of the output array. It is recommended to provide this. If not provided, will be inferred by applying the function to a small set of fake data.

chunks : tuple, optional

Chunk shape of resulting blocks if the function does not preserve shape. If not provided, the resulting array is assumed to have the same block structure as the first input array.

drop_axis : number or iterable, optional

Dimensions lost by the function.

new_axis : number or iterable, optional

New dimensions created by the function. Note that these are applied after drop_axis (if present).

token : string, optional

The key prefix to use for the output array. If not provided, will be determined from the function name.

name : string, optional

The key name to use for the output array. Note that this fully specifies the output key name, and must be unique. If not provided, will be determined by a hash of the arguments.

**kwargs :

Other keyword arguments to pass to function. Values must be constants (not dask.arrays)

Examples

>>> import dask.array as da
>>> x = da.arange(6, chunks=3)

>>> x.map_blocks(lambda x: x * 2).compute()
array([ 0,  2,  4,  6,  8, 10])

The da.map_blocks function can also accept multiple arrays.

>>> d = da.arange(5, chunks=2)
>>> e = da.arange(5, chunks=2)

>>> f = map_blocks(lambda a, b: a + b**2, d, e)
>>> f.compute()
array([ 0,  2,  6, 12, 20])

If the function changes shape of the blocks then you must provide chunks explicitly.

>>> y = x.map_blocks(lambda x: x[::2], chunks=((2, 2),))

You have a bit of freedom in specifying chunks. If all of the output chunk sizes are the same, you can provide just that chunk size as a single tuple.

>>> a = da.arange(18, chunks=(6,))
>>> b = a.map_blocks(lambda x: x[:3], chunks=(3,))

If the function changes the dimension of the blocks you must specify the created or destroyed dimensions.

>>> b = a.map_blocks(lambda x: x[None, :, None], chunks=(1, 6, 1),
...                  new_axis=[0, 2])

Map_blocks aligns blocks by block positions without regard to shape. In the following example we have two arrays with the same number of blocks but with different shape and chunk sizes.

>>> x = da.arange(1000, chunks=(100,))
>>> y = da.arange(100, chunks=(10,))

The relevant attribute to match is numblocks.

>>> x.numblocks
(10,)
>>> y.numblocks
(10,)

If these match (up to broadcasting rules) then we can map arbitrary functions across blocks

>>> def func(a, b):
...     return np.array([a.max(), b.max()])

>>> da.map_blocks(func, x, y, chunks=(2,), dtype='i8')
dask.array<func, shape=(20,), dtype=int64, chunksize=(2,)>

>>> _.compute()
array([ 99,   9, 199,  19, 299,  29, 399,  39, 499,  49, 599,  59, 699,
        69, 799,  79, 899,  89, 999,  99])

Your block function can learn where in the array it is if it supports a block_id keyword argument. This will receive entries like (2, 0, 1), the position of the block in the dask array.

>>> def func(block, block_id=None):
...     pass

You may specify the key name prefix of the resulting task in the graph with the optional token keyword argument.

>>> x.map_blocks(lambda x: x + 1, token='increment')  
dask.array<increment, shape=(100,), dtype=int64, chunksize=(10,)>
map_overlap(func, depth, boundary=None, trim=True, **kwargs)

Map a function over blocks of the array with some overlap

We share neighboring zones between blocks of the array, then map a function, then trim away the neighboring strips.

Parameters:

func: function

The function to apply to each extended block

depth: int, tuple, or dict

The number of cells that each block should share with its neighbors If a tuple or dict this can be different per axis

boundary: str, tuple, dict

how to handle the boundaries. Values include ‘reflect’, ‘periodic’, ‘nearest’, ‘none’, or any constant value like 0 or np.nan

trim: bool

Whether or not to trim the excess after the map function. Set this to false if your mapping function does this for you.

**kwargs:

Other keyword arguments valid in map_blocks

Examples

>>> x = np.array([1, 1, 2, 3, 3, 3, 2, 1, 1])
>>> x = from_array(x, chunks=5)
>>> def derivative(x):
...     return x - np.roll(x, 1)

>>> y = x.map_overlap(derivative, depth=1, boundary=0)
>>> y.compute()
array([ 1,  0,  1,  1,  0,  0, -1, -1,  0])

>>> import dask.array as da
>>> x = np.arange(16).reshape((4, 4))
>>> d = da.from_array(x, chunks=(2, 2))
>>> d.map_overlap(lambda x: x + x.size, depth=1).compute()
array([[16, 17, 18, 19],
       [20, 21, 22, 23],
       [24, 25, 26, 27],
       [28, 29, 30, 31]])

>>> func = lambda x: x + x.size
>>> depth = {0: 1, 1: 1}
>>> boundary = {0: 'reflect', 1: 'none'}
>>> d.map_overlap(func, depth, boundary).compute()  
array([[12,  13,  14,  15],
       [16,  17,  18,  19],
       [20,  21,  22,  23],
       [24,  25,  26,  27]])
max(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the maximum of an array or maximum along an axis.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

Axis or axes along which to operate. By default, flattened input is used.

New in version 1.7.0.

If this is a tuple of ints, the maximum is selected over multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternative output array in which to place the result. Must be of the same shape and buffer length as the expected output. See doc.ufuncs (Section “Output arguments”) for more details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the amax method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

amax : ndarray or scalar

Maximum of a. If axis is None, the result is a scalar value. If axis is given, the result is an array of dimension a.ndim - 1.

See also

amin
The minimum value of an array along a given axis, propagating any NaNs.
nanmax
The maximum value of an array along a given axis, ignoring any NaNs.
maximum
Element-wise maximum of two arrays, propagating any NaNs.
fmax
Element-wise maximum of two arrays, ignoring any NaNs.
argmax
Return the indices of the maximum values.

nanmin, minimum, fmin

Notes

NaN values are propagated, that is if at least one item is NaN, the corresponding max value will be NaN as well. To ignore NaN values (MATLAB behavior), please use nanmax.

Don’t use amax for element-wise comparison of 2 arrays; when a.shape[0] is 2, maximum(a[0], a[1]) is faster than amax(a, axis=0).

Examples

>>> a = np.arange(4).reshape((2,2))
>>> a
array([[0, 1],
       [2, 3]])
>>> np.amax(a)           # Maximum of the flattened array
3
>>> np.amax(a, axis=0)   # Maxima along the first axis
array([2, 3])
>>> np.amax(a, axis=1)   # Maxima along the second axis
array([1, 3])

>>> b = np.arange(5, dtype=np.float)
>>> b[2] = np.NaN
>>> np.amax(b)
nan
>>> np.nanmax(b)
4.0
mean(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Compute the arithmetic mean along the specified axis.

Returns the average of the array elements. The average is taken over the flattened array by default, otherwise over the specified axis. float64 intermediate and return values are used for integer inputs.

Parameters:

a : array_like

Array containing numbers whose mean is desired. If a is not an array, a conversion is attempted.

axis : None or int or tuple of ints, optional

Axis or axes along which the means are computed. The default is to compute the mean of the flattened array.

New in version 1.7.0.

If this is a tuple of ints, a mean is performed over multiple axes, instead of a single axis or all the axes as before.

dtype : data-type, optional

Type to use in computing the mean. For integer inputs, the default is float64; for floating point inputs, it is the same as the input dtype.

out : ndarray, optional

Alternate output array in which to place the result. The default is None; if provided, it must have the same shape as the expected output, but the type will be cast if necessary. See doc.ufuncs for details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the mean method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

m : ndarray, see dtype parameter above

If out=None, returns a new array containing the mean values, otherwise a reference to the output array is returned.

See also

average
Weighted average

std, var, nanmean, nanstd, nanvar

Notes

The arithmetic mean is the sum of the elements along the axis divided by the number of elements.

Note that for floating-point input, the mean is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-precision accumulator using the dtype keyword can alleviate this issue.

By default, float16 results are computed using float32 intermediates for extra precision.

Examples

>>> a = np.array([[1, 2], [3, 4]])
>>> np.mean(a)
2.5
>>> np.mean(a, axis=0)
array([ 2.,  3.])
>>> np.mean(a, axis=1)
array([ 1.5,  3.5])

In single precision, mean can be inaccurate:

>>> a = np.zeros((2, 512*512), dtype=np.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.mean(a)
0.54999924

Computing the mean in float64 is more accurate:

>>> np.mean(a, dtype=np.float64)
0.55000000074505806
min(a, axis=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the minimum of an array or minimum along an axis.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

Axis or axes along which to operate. By default, flattened input is used.

New in version 1.7.0.

If this is a tuple of ints, the minimum is selected over multiple axes, instead of a single axis or all the axes as before.

out : ndarray, optional

Alternative output array in which to place the result. Must be of the same shape and buffer length as the expected output. See doc.ufuncs (Section “Output arguments”) for more details.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the amin method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

amin : ndarray or scalar

Minimum of a. If axis is None, the result is a scalar value. If axis is given, the result is an array of dimension a.ndim - 1.

See also

amax
The maximum value of an array along a given axis, propagating any NaNs.
nanmin
The minimum value of an array along a given axis, ignoring any NaNs.
minimum
Element-wise minimum of two arrays, propagating any NaNs.
fmin
Element-wise minimum of two arrays, ignoring any NaNs.
argmin
Return the indices of the minimum values.

nanmax, maximum, fmax

Notes

NaN values are propagated, that is if at least one item is NaN, the corresponding min value will be NaN as well. To ignore NaN values (MATLAB behavior), please use nanmin.

Don’t use amin for element-wise comparison of 2 arrays; when a.shape[0] is 2, minimum(a[0], a[1]) is faster than amin(a, axis=0).

Examples

>>> a = np.arange(4).reshape((2,2))
>>> a
array([[0, 1],
       [2, 3]])
>>> np.amin(a)           # Minimum of the flattened array
0
>>> np.amin(a, axis=0)   # Minima along the first axis
array([0, 1])
>>> np.amin(a, axis=1)   # Minima along the second axis
array([0, 2])

>>> b = np.arange(5, dtype=np.float)
>>> b[2] = np.NaN
>>> np.amin(b)
nan
>>> np.nanmin(b)
0.0
moment(order, axis=None, dtype=None, keepdims=False, ddof=0, split_every=None, out=None)

Calculate the nth centralized moment.

Parameters:

order : int

Order of the moment that is returned, must be >= 2.

axis : int, optional

Axis along which the central moment is computed. The default is to compute the moment of the flattened array.

dtype : data-type, optional

Type to use in computing the moment. For arrays of integer type the default is float64; for arrays of float types it is the same as the array type.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the original array.

ddof : int, optional

“Delta Degrees of Freedom”: the divisor used in the calculation is N - ddof, where N represents the number of elements. By default ddof is zero.
Returns:

moment : ndarray

References

[R413413414]Pebay, Philippe (2008), “Formulas for Robust, One-Pass Parallel

Computation of Covariances and Arbitrary-Order Statistical Moments” (PDF), Technical Report SAND2008-6212, Sandia National Laboratories

nbytes

Number of bytes in array

prod(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Return the product of array elements over a given axis.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

Axis or axes along which a product is performed. The default, axis=None, will calculate the product of all the elements in the input array. If axis is negative it counts from the last to the first axis.

New in version 1.7.0.

If axis is a tuple of ints, a product is performed on all of the axes specified in the tuple instead of a single axis or all the axes as before.

dtype : dtype, optional

The type of the returned array, as well as of the accumulator in which the elements are multiplied. The dtype of a is used by default unless a has an integer dtype of less precision than the default platform integer. In that case, if a is signed then the platform integer is used while if a is unsigned then an unsigned integer of the same precision as the platform integer is used.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the prod method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

product_along_axis : ndarray, see dtype parameter above.

An array shaped as a but with the specified axis removed. Returns a reference to out if specified.

See also

ndarray.prod
equivalent method
numpy.doc.ufuncs
Section “Output arguments”

Notes

Arithmetic is modular when using integer types, and no error is raised on overflow. That means that, on a 32-bit platform:

>>> x = np.array([536870910, 536870910, 536870910, 536870910])
>>> np.prod(x) #random
16

The product of an empty array is the neutral element 1:

>>> np.prod([])
1.0

Examples

By default, calculate the product of all elements:

>>> np.prod([1.,2.])
2.0

Even when the input array is two-dimensional:

>>> np.prod([[1.,2.],[3.,4.]])
24.0

But we can also specify the axis over which to multiply:

>>> np.prod([[1.,2.],[3.,4.]], axis=1)
array([  2.,  12.])

If the type of x is unsigned, then the output type is the unsigned platform integer:

>>> x = np.array([1, 2, 3], dtype=np.uint8)
>>> np.prod(x).dtype == np.uint
True

If x is of a signed integer type, then the output type is the default platform integer:

>>> x = np.array([1, 2, 3], dtype=np.int8)
>>> np.prod(x).dtype == np.int
True
ravel(a, order='C')

Return a contiguous flattened array.

A 1-D array, containing the elements of the input, is returned. A copy is made only if needed.

As of NumPy 1.10, the returned array will have the same type as the input array. (for example, a masked array will be returned for a masked array input)

Parameters:

a : array_like

Input array. The elements in a are read in the order specified by order, and packed as a 1-D array.

order : {‘C’,’F’, ‘A’, ‘K’}, optional

The elements of a are read using this index order. ‘C’ means to index the elements in row-major, C-style order, with the last axis index changing fastest, back to the first axis index changing slowest. ‘F’ means to index the elements in column-major, Fortran-style order, with the first index changing fastest, and the last index changing slowest. Note that the ‘C’ and ‘F’ options take no account of the memory layout of the underlying array, and only refer to the order of axis indexing. ‘A’ means to read the elements in Fortran-like index order if a is Fortran contiguous in memory, C-like order otherwise. ‘K’ means to read the elements in the order they occur in memory, except for reversing the data when strides are negative. By default, ‘C’ index order is used.
Returns:

y : array_like

If a is a matrix, y is a 1-D ndarray, otherwise y is an array of the same subtype as a. The shape of the returned array is (a.size,). Matrices are special cased for backward compatibility.

See also

ndarray.flat
1-D iterator over an array.
ndarray.flatten
1-D array copy of the elements of an array in row-major order.
ndarray.reshape
Change the shape of an array without changing its data.

Notes

In row-major, C-style order, in two dimensions, the row index varies the slowest, and the column index the quickest. This can be generalized to multiple dimensions, where row-major order implies that the index along the first axis varies slowest, and the index along the last quickest. The opposite holds for column-major, Fortran-style index ordering.

When a view is desired in as many cases as possible, arr.reshape(-1) may be preferable.

Examples

It is equivalent to reshape(-1, order=order).

>>> x = np.array([[1, 2, 3], [4, 5, 6]])
>>> print(np.ravel(x))
[1 2 3 4 5 6]

>>> print(x.reshape(-1))
[1 2 3 4 5 6]

>>> print(np.ravel(x, order='F'))
[1 4 2 5 3 6]

When order is ‘A’, it will preserve the array’s ‘C’ or ‘F’ ordering:

>>> print(np.ravel(x.T))
[1 4 2 5 3 6]
>>> print(np.ravel(x.T, order='A'))
[1 2 3 4 5 6]

When order is ‘K’, it will preserve orderings that are neither ‘C’ nor ‘F’, but won’t reverse axes:

>>> a = np.arange(3)[::-1]; a
array([2, 1, 0])
>>> a.ravel(order='C')
array([2, 1, 0])
>>> a.ravel(order='K')
array([2, 1, 0])

>>> a = np.arange(12).reshape(2,3,2).swapaxes(1,2); a
array([[[ 0,  2,  4],
        [ 1,  3,  5]],
       [[ 6,  8, 10],
        [ 7,  9, 11]]])
>>> a.ravel(order='C')
array([ 0,  2,  4,  1,  3,  5,  6,  8, 10,  7,  9, 11])
>>> a.ravel(order='K')
array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11])
rechunk(chunks, threshold=None, block_size_limit=None)

See da.rechunk for docstring

repeat(a, repeats, axis=None)

Repeat elements of an array.

Parameters:

a : array_like

Input array.

repeats : int or array of ints

The number of repetitions for each element. repeats is broadcasted to fit the shape of the given axis.

axis : int, optional

The axis along which to repeat values. By default, use the flattened input array, and return a flat output array.
Returns:

repeated_array : ndarray

Output array which has the same shape as a, except along the given axis.

See also

tile
Tile an array.

Examples

>>> np.repeat(3, 4)
array([3, 3, 3, 3])
>>> x = np.array([[1,2],[3,4]])
>>> np.repeat(x, 2)
array([1, 1, 2, 2, 3, 3, 4, 4])
>>> np.repeat(x, 3, axis=1)
array([[1, 1, 1, 2, 2, 2],
       [3, 3, 3, 4, 4, 4]])
>>> np.repeat(x, [1, 2], axis=0)
array([[1, 2],
       [3, 4],
       [3, 4]])
reshape(a, newshape, order='C')

Gives a new shape to an array without changing its data.

Parameters:

a : array_like

Array to be reshaped.

newshape : int or tuple of ints

The new shape should be compatible with the original shape. If an integer, then the result will be a 1-D array of that length. One shape dimension can be -1. In this case, the value is inferred from the length of the array and remaining dimensions.

order : {‘C’, ‘F’, ‘A’}, optional

Read the elements of a using this index order, and place the elements into the reshaped array using this index order. ‘C’ means to read / write the elements using C-like index order, with the last axis index changing fastest, back to the first axis index changing slowest. ‘F’ means to read / write the elements using Fortran-like index order, with the first index changing fastest, and the last index changing slowest. Note that the ‘C’ and ‘F’ options take no account of the memory layout of the underlying array, and only refer to the order of indexing. ‘A’ means to read / write the elements in Fortran-like index order if a is Fortran contiguous in memory, C-like order otherwise.
Returns:

reshaped_array : ndarray

This will be a new view object if possible; otherwise, it will be a copy. Note there is no guarantee of the memory layout (C- or Fortran- contiguous) of the returned array.

See also

ndarray.reshape
Equivalent method.

Notes

It is not always possible to change the shape of an array without copying the data. If you want an error to be raise if the data is copied, you should assign the new shape to the shape attribute of the array:

>>> a = np.zeros((10, 2))
# A transpose make the array non-contiguous
>>> b = a.T
# Taking a view makes it possible to modify the shape without modifying
# the initial object.
>>> c = b.view()
>>> c.shape = (20)
AttributeError: incompatible shape for a non-contiguous array

The order keyword gives the index ordering both for fetching the values from a, and then placing the values into the output array. For example, let’s say you have an array:

>>> a = np.arange(6).reshape((3, 2))
>>> a
array([[0, 1],
       [2, 3],
       [4, 5]])

You can think of reshaping as first raveling the array (using the given index order), then inserting the elements from the raveled array into the new array using the same kind of index ordering as was used for the raveling.

>>> np.reshape(a, (2, 3)) # C-like index ordering
array([[0, 1, 2],
       [3, 4, 5]])
>>> np.reshape(np.ravel(a), (2, 3)) # equivalent to C ravel then C reshape
array([[0, 1, 2],
       [3, 4, 5]])
>>> np.reshape(a, (2, 3), order='F') # Fortran-like index ordering
array([[0, 4, 3],
       [2, 1, 5]])
>>> np.reshape(np.ravel(a, order='F'), (2, 3), order='F')
array([[0, 4, 3],
       [2, 1, 5]])

Examples

>>> a = np.array([[1,2,3], [4,5,6]])
>>> np.reshape(a, 6)
array([1, 2, 3, 4, 5, 6])
>>> np.reshape(a, 6, order='F')
array([1, 4, 2, 5, 3, 6])

>>> np.reshape(a, (3,-1))       # the unspecified value is inferred to be 2
array([[1, 2],
       [3, 4],
       [5, 6]])
round(a, decimals=0, out=None)

Round an array to the given number of decimals.

Refer to around for full documentation.

See also

around
equivalent function
size

Number of elements in array

squeeze(a, axis=None)

Remove single-dimensional entries from the shape of an array.

Parameters:

a : array_like

Input data.

axis : None or int or tuple of ints, optional

New in version 1.7.0.

Selects a subset of the single-dimensional entries in the shape. If an axis is selected with shape entry greater than one, an error is raised.
Returns:

squeezed : ndarray

The input array, but with all or a subset of the dimensions of length 1 removed. This is always a itself or a view into a.

Examples

>>> x = np.array([[[0], [1], [2]]])
>>> x.shape
(1, 3, 1)
>>> np.squeeze(x).shape
(3,)
>>> np.squeeze(x, axis=(2,)).shape
(1, 3)
std(a, axis=None, dtype=None, out=None, ddof=0, keepdims=<class 'numpy._globals._NoValue'>)

Compute the standard deviation along the specified axis.

Returns the standard deviation, a measure of the spread of a distribution, of the array elements. The standard deviation is computed for the flattened array by default, otherwise over the specified axis.

Parameters:

a : array_like

Calculate the standard deviation of these values.

axis : None or int or tuple of ints, optional

Axis or axes along which the standard deviation is computed. The default is to compute the standard deviation of the flattened array.

New in version 1.7.0.

If this is a tuple of ints, a standard deviation is performed over multiple axes, instead of a single axis or all the axes as before.

dtype : dtype, optional

Type to use in computing the standard deviation. For arrays of integer type the default is float64, for arrays of float types it is the same as the array type.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output but the type (of the calculated values) will be cast if necessary.

ddof : int, optional

Means Delta Degrees of Freedom. The divisor used in calculations is N - ddof, where N represents the number of elements. By default ddof is zero.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the std method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

standard_deviation : ndarray, see dtype parameter above.

If out is None, return a new array containing the standard deviation, otherwise return a reference to the output array.

See also

var, mean, nanmean, nanstd, nanvar

numpy.doc.ufuncs
Section “Output arguments”

Notes

The standard deviation is the square root of the average of the squared deviations from the mean, i.e., std = sqrt(mean(abs(x - x.mean())**2)).

The average squared deviation is normally calculated as x.sum() / N, where N = len(x). If, however, ddof is specified, the divisor N - ddof is used instead. In standard statistical practice, ddof=1 provides an unbiased estimator of the variance of the infinite population. ddof=0 provides a maximum likelihood estimate of the variance for normally distributed variables. The standard deviation computed in this function is the square root of the estimated variance, so even with ddof=1, it will not be an unbiased estimate of the standard deviation per se.

Note that, for complex numbers, std takes the absolute value before squaring, so that the result is always real and nonnegative.

For floating-point input, the std is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-accuracy accumulator using the dtype keyword can alleviate this issue.

Examples

>>> a = np.array([[1, 2], [3, 4]])
>>> np.std(a)
1.1180339887498949
>>> np.std(a, axis=0)
array([ 1.,  1.])
>>> np.std(a, axis=1)
array([ 0.5,  0.5])

In single precision, std() can be inaccurate:

>>> a = np.zeros((2, 512*512), dtype=np.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.std(a)
0.45000005

Computing the standard deviation in float64 is more accurate:

>>> np.std(a, dtype=np.float64)
0.44999999925494177
store(sources, targets, lock=True, regions=None, compute=True, **kwargs)

Store dask arrays in array-like objects, overwrite data in target

This stores dask arrays into object that supports numpy-style setitem indexing. It stores values chunk by chunk so that it does not have to fill up memory. For best performance you can align the block size of the storage target with the block size of your array.

If your data fits in memory then you may prefer calling np.array(myarray) instead.

Parameters:

sources: Array or iterable of Arrays

targets: array-like or iterable of array-likes

These should support setitem syntax target[10:20] = ...

lock: boolean or threading.Lock, optional

Whether or not to lock the data stores while storing. Pass True (lock each file individually), False (don’t lock) or a particular threading.Lock object to be shared among all writes.

regions: tuple of slices or iterable of tuple of slices

Each region tuple in regions should be such that target[region].shape = source.shape for the corresponding source and target in sources and targets, respectively.

compute: boolean, optional

If true compute immediately, return dask.delayed.Delayed otherwise

Examples

>>> x = ...  

>>> import h5py  
>>> f = h5py.File('myfile.hdf5')  
>>> dset = f.create_dataset('/data', shape=x.shape,
...                                  chunks=x.chunks,
...                                  dtype='f8')  

>>> store(x, dset)

Alternatively store many arrays at the same time

>>> store([x, y, z], [dset1, dset2, dset3])
sum(a, axis=None, dtype=None, out=None, keepdims=<class 'numpy._globals._NoValue'>)

Sum of array elements over a given axis.

Parameters:

a : array_like

Elements to sum.

axis : None or int or tuple of ints, optional

Axis or axes along which a sum is performed. The default, axis=None, will sum all of the elements of the input array. If axis is negative it counts from the last to the first axis.

New in version 1.7.0.

If axis is a tuple of ints, a sum is performed on all of the axes specified in the tuple instead of a single axis or all the axes as before.

dtype : dtype, optional

The type of the returned array and of the accumulator in which the elements are summed. The dtype of a is used by default unless a has an integer dtype of less precision than the default platform integer. In that case, if a is signed then the platform integer is used while if a is unsigned then an unsigned integer of the same precision as the platform integer is used.

out : ndarray, optional

Alternative output array in which to place the result. It must have the same shape as the expected output, but the type of the output values will be cast if necessary.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the sum method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

sum_along_axis : ndarray

An array with the same shape as a, with the specified axis removed. If a is a 0-d array, or if axis is None, a scalar is returned. If an output array is specified, a reference to out is returned.

See also

ndarray.sum
Equivalent method.
cumsum
Cumulative sum of array elements.
trapz
Integration of array values using the composite trapezoidal rule.

mean, average

Notes

Arithmetic is modular when using integer types, and no error is raised on overflow.

The sum of an empty array is the neutral element 0:

>>> np.sum([])
0.0

Examples

>>> np.sum([0.5, 1.5])
2.0
>>> np.sum([0.5, 0.7, 0.2, 1.5], dtype=np.int32)
1
>>> np.sum([[0, 1], [0, 5]])
6
>>> np.sum([[0, 1], [0, 5]], axis=0)
array([0, 6])
>>> np.sum([[0, 1], [0, 5]], axis=1)
array([1, 5])

If the accumulator is too small, overflow occurs:

>>> np.ones(128, dtype=np.int8).sum(dtype=np.int8)
-128
swapaxes(a, axis1, axis2)

Interchange two axes of an array.

Parameters:

a : array_like

Input array.

axis1 : int

First axis.

axis2 : int

Second axis.
Returns:

a_swapped : ndarray

For NumPy >= 1.10.0, if a is an ndarray, then a view of a is returned; otherwise a new array is created. For earlier NumPy versions a view of a is returned only if the order of the axes is changed, otherwise the input array is returned.

Examples

>>> x = np.array([[1,2,3]])
>>> np.swapaxes(x,0,1)
array([[1],
       [2],
       [3]])

>>> x = np.array([[[0,1],[2,3]],[[4,5],[6,7]]])
>>> x
array([[[0, 1],
        [2, 3]],
       [[4, 5],
        [6, 7]]])

>>> np.swapaxes(x,0,2)
array([[[0, 4],
        [2, 6]],
       [[1, 5],
        [3, 7]]])
to_dask_dataframe(columns=None)

Convert dask Array to dask Dataframe

Parameters:

columns: list or string

list of column names if DataFrame, single string if Series

See also

dask.dataframe.from_dask_array

to_delayed()

Convert Array into dask Delayed objects

Returns an array of values, one value per chunk.

See also

dask.array.from_delayed

to_hdf5(filename, datapath, **kwargs)

Store array in HDF5 file

>>> x.to_hdf5('myfile.hdf5', '/x')

Optionally provide arguments as though to h5py.File.create_dataset

>>> x.to_hdf5('myfile.hdf5', '/x', compression='lzf', shuffle=True)

See also

da.store, h5py.File.create_dataset

topk(k, x)

The top k elements of an array

Returns the k greatest elements of the array in sorted order. Only works on arrays of a single dimension.

This assumes that k is small. All results will be returned in a single chunk.

Examples

>>> x = np.array([5, 1, 3, 6])
>>> d = from_array(x, chunks=2)
>>> d.topk(2).compute()
array([6, 5])
transpose(*axes)

Returns a view of the array with axes transposed.

For a 1-D array, this has no effect. (To change between column and row vectors, first cast the 1-D array into a matrix object.) For a 2-D array, this is the usual matrix transpose. For an n-D array, if axes are given, their order indicates how the axes are permuted (see Examples). If axes are not provided and a.shape = (i[0], i[1], ... i[n-2], i[n-1]), then a.transpose().shape = (i[n-1], i[n-2], ... i[1], i[0]).

Parameters:

axes : None, tuple of ints, or n ints

  • None or no argument: reverses the order of the axes.
  • tuple of ints: i in the j-th place in the tuple means a‘s i-th axis becomes a.transpose()‘s j-th axis.
  • n ints: same as an n-tuple of the same ints (this form is intended simply as a “convenience” alternative to the tuple form)
Returns:

out : ndarray

View of a, with axes suitably permuted.

See also

ndarray.T
Array property returning the array transposed.

Examples

>>> a = np.array([[1, 2], [3, 4]])  
>>> a  
array([[1, 2],
       [3, 4]])
>>> a.transpose()  
array([[1, 3],
       [2, 4]])
>>> a.transpose((1, 0))  
array([[1, 3],
       [2, 4]])
>>> a.transpose(1, 0)  
array([[1, 3],
       [2, 4]])
var(a, axis=None, dtype=None, out=None, ddof=0, keepdims=<class 'numpy._globals._NoValue'>)

Compute the variance along the specified axis.

Returns the variance of the array elements, a measure of the spread of a distribution. The variance is computed for the flattened array by default, otherwise over the specified axis.

Parameters:

a : array_like

Array containing numbers whose variance is desired. If a is not an array, a conversion is attempted.

axis : None or int or tuple of ints, optional

Axis or axes along which the variance is computed. The default is to compute the variance of the flattened array.

New in version 1.7.0.

If this is a tuple of ints, a variance is performed over multiple axes, instead of a single axis or all the axes as before.

dtype : data-type, optional

Type to use in computing the variance. For arrays of integer type the default is float32; for arrays of float types it is the same as the array type.

out : ndarray, optional

Alternate output array in which to place the result. It must have the same shape as the expected output, but the type is cast if necessary.

ddof : int, optional

“Delta Degrees of Freedom”: the divisor used in the calculation is N - ddof, where N represents the number of elements. By default ddof is zero.

keepdims : bool, optional

If this is set to True, the axes which are reduced are left in the result as dimensions with size one. With this option, the result will broadcast correctly against the input array.

If the default value is passed, then keepdims will not be passed through to the var method of sub-classes of ndarray, however any non-default value will be. If the sub-classes sum method does not implement keepdims any exceptions will be raised.
Returns:

variance : ndarray, see dtype parameter above

If out=None, returns a new array containing the variance; otherwise, a reference to the output array is returned.

See also

std, mean, nanmean, nanstd, nanvar

numpy.doc.ufuncs
Section “Output arguments”

Notes

The variance is the average of the squared deviations from the mean, i.e., var = mean(abs(x - x.mean())**2).

The mean is normally calculated as x.sum() / N, where N = len(x). If, however, ddof is specified, the divisor N - ddof is used instead. In standard statistical practice, ddof=1 provides an unbiased estimator of the variance of a hypothetical infinite population. ddof=0 provides a maximum likelihood estimate of the variance for normally distributed variables.

Note that for complex numbers, the absolute value is taken before squaring, so that the result is always real and nonnegative.

For floating-point input, the variance is computed using the same precision the input has. Depending on the input data, this can cause the results to be inaccurate, especially for float32 (see example below). Specifying a higher-accuracy accumulator using the dtype keyword can alleviate this issue.

Examples

>>> a = np.array([[1, 2], [3, 4]])
>>> np.var(a)
1.25
>>> np.var(a, axis=0)
array([ 1.,  1.])
>>> np.var(a, axis=1)
array([ 0.25,  0.25])

In single precision, var() can be inaccurate:

>>> a = np.zeros((2, 512*512), dtype=np.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> np.var(a)
0.20250003

Computing the variance in float64 is more accurate:

>>> np.var(a, dtype=np.float64)
0.20249999932944759
>>> ((1-0.55)**2 + (0.1-0.55)**2)/2
0.2025
view(dtype, order='C')

Get a view of the array as a new data type

Parameters:

dtype:

The dtype by which to view the array

order: string

‘C’ or ‘F’ (Fortran) ordering

This reinterprets the bytes of the array under a new dtype. If that

dtype does not have the same size as the original array then the shape

will change.

Beware that both numpy and dask.array can behave oddly when taking

shape-changing views of arrays under Fortran ordering. Under some

versions of NumPy this function will fail when taking shape-changing

views of Fortran ordered arrays if the first dimension has chunks of

size one.
vnorm(ord=None, axis=None, keepdims=False, split_every=None, out=None)

Vector norm