jax.random.categorical

jax.random.categorical(key, logits, axis=-1, shape=None, replace=True)

Sample random values from categorical distributions.

Sampling with replacement uses the Gumbel max trick. Sampling without replacement uses the Gumbel top-k trick. See [1] for reference.

Parameters:

• key (ArrayLike) – a PRNG key used as the random key.

• logits (RealArray) – Unnormalized log probabilities of the categorical distribution(s) to sample from, so that softmax(logits, axis) gives the corresponding probabilities.

• axis (int) – Axis along which logits belong to the same categorical distribution.

• shape (Shape | None | None) – Optional, a tuple of nonnegative integers representing the result shape. Must be broadcast-compatible with np.delete(logits.shape, axis). The default (None) produces a result shape equal to np.delete(logits.shape, axis).

• replace (bool) – If True (default), perform sampling with replacement. If False, perform sampling without replacement.

Returns:

A random array with int dtype and shape given by shape if shape is not None, or else np.delete(logits.shape, axis).

Return type:

Array

References

[1]

Wouter Kool, Herke van Hoof, Max Welling. “Stochastic Beams and Where to Find Them: The Gumbel-Top-k Trick for Sampling Sequences Without Replacement”. Proceedings of the 36th International Conference on Machine Learning, PMLR 97:3499-3508, 2019. https://proceedings.mlr.press/v97/kool19a.html.