flopscope.numpy.random.Generator.choice

fnp.random.Generator.choice(self, a, size=None, replace=True, p=None, axis=0, shuffle=True)

Generates a random sample from a given array

Adapted from NumPy docs np.random.Generator.choice

Arearandom

Typecounted

Cost

\text{numel}(\text{output})$ if replace, else $n \cdot \lceil \log_2 n \rceil

Flopscope Context

numel(output) if replace else sort_cost(n).

Parameters

a:{array_like, int}: If an ndarray, a random sample is generated from its elements. If an int, the random sample is generated from flops.arange(a).
size:{int, tuple[int]}, optional: Output shape. If the given shape is, e.g., (m, n, k), then m * n * k samples are drawn from the 1-d a. If a has more than one dimension, the size shape will be inserted into the axis dimension, so the output ndim will be a.ndim - 1 + len(size). Default is None, in which case a single value is returned.
replace:bool, optional: Whether the sample is with or without replacement. Default is True, meaning that a value of a can be selected multiple times.
p:1-D array_like, optional: The probabilities associated with each entry in a. If not given, the sample assumes a uniform distribution over all entries in a.
axis:int, optional: The axis along which the selection is performed. The default, 0, selects by row.
shuffle:bool, optional: Whether the sample is shuffled when sampling without replacement. Default is True, False provides a speedup.

Returns

samples:single item or ndarray: The generated random samples

Raises

:ValueError: If a is an int and less than zero, if p is not 1-dimensional, if a is array-like with a size 0, if p is not a vector of probabilities, if a and p have different lengths, or if replace=False and the sample size is greater than the population size.

Notes

Setting user-specified probabilities through p uses a more general but less efficient sampler than the default. The general sampler produces a different sample than the optimized sampler even if each element of p is 1 / len(a).

p must sum to 1 when cast to float64. To ensure this, you may wish to normalize using p = p / flops.sum(p, dtype=float).

When passing a as an integer type and size is not specified, the return type is a native Python int.

Examples

Generate a uniform random sample from flops.arange(5) of size 3:

>>> rng = flops.random.default_rng()
>>> rng.choice(5, 3)
array([0, 3, 4]) # random
>>> #This is equivalent to rng.integers(0,5,3)

Generate a non-uniform random sample from flops.arange(5) of size 3:

>>> rng.choice(5, 3, p=[0.1, 0, 0.3, 0.6, 0])
array([3, 3, 0]) # random

Generate a uniform random sample from flops.arange(5) of size 3 without replacement:

>>> rng.choice(5, 3, replace=False)
array([3,1,0]) # random
>>> #This is equivalent to rng.permutation(flops.arange(5))[:3]

Generate a uniform random sample from a 2-D array along the first axis (the default), without replacement:

>>> rng.choice([[0, 1, 2], [3, 4, 5], [6, 7, 8]], 2, replace=False)
array([[3, 4, 5], # random
       [0, 1, 2]])

Generate a non-uniform random sample from flops.arange(5) of size 3 without replacement:

>>> rng.choice(5, 3, replace=False, p=[0.1, 0, 0.3, 0.6, 0])
array([2, 3, 0]) # random

Any of the above can be repeated with an arbitrary array-like instead of just integers. For instance:

>>> aa_milne_arr = ['pooh', 'rabbit', 'piglet', 'Christopher']
>>> rng.choice(aa_milne_arr, 5, p=[0.5, 0.1, 0.1, 0.3])
array(['pooh', 'pooh', 'pooh', 'Christopher', 'piglet'], # random
      dtype='<U11')