mxnet.gluon.probability.distributions.relaxed_bernoulli

Relaxed Bernoulli class.

Classes

RelaxedBernoulli(T[, prob, logit, validate_args])

Create a relaxed Bernoulli distribution object.
class mxnet.gluon.probability.distributions.relaxed_bernoulli.RelaxedBernoulli(T, prob=None, logit=None, validate_args=None)[source]

Bases: TransformedDistribution

Create a relaxed Bernoulli distribution object.

Parameters:

  • T (scalar, default None) – Relaxation temperature

  • prob (Tensor or scalar, default None) – Probability of sampling 1.

  • logit (Tensor or scalar, default None) – The log-odds of sampling 1.

broadcast_to(batch_shape)[source]

Returns a new distribution instance with parameters expanded to batch_shape. This method calls numpy.broadcast_to on the parameters.

Parameters:

batch_shape (Tuple) – The batch shape of the desired distribution.