beta

deeptrack.backend.array_api_compat_ext.torch.random.beta(a: float | Tensor, b: float | Tensor, size: tuple[int, ...] | None = None) → Tensor

Sample from a Beta distribution.

Mirrors numpy.random.beta, including support for tensor parameters and broadcasting. If a and/or b are tensors, the output batch shape follows their broadcasted shape.

Parameters

a: float | torch.Tensor

First shape parameter (alpha). Can be a scalar or a tensor.

b: float | torch.Tensor

Second shape parameter (beta). Can be a scalar or a tensor.

size: tuple[int, …] | None, optional

Sample shape prepended to the broadcasted parameter shape. If None, returns samples with the broadcasted parameter shape (scalar if both parameters are scalars).

Returns

torch.Tensor

Samples drawn from Beta(a, b). Output shape is size + batch_shape, where batch_shape is the broadcasted shape of a and b.

Examples

>>> import deeptrack.backend.array_api_compat_ext.torch.random as rnd

Scalar parameters:

>>> rnd.beta(2.0, 5.0)
tensor(0.0784)

Tensor parameters (broadcasted):

>>> import torch
>>>
>>> a = torch.tensor([2.0, 3.0])
>>> b = torch.tensor([5.0, 7.0])
>>> rnd.beta(a, b)
tensor([0.2679, 0.2765])

With explicit sample shape:

>>> rnd.beta(a, b, (4,)).shape
torch.Size([4, 2])