mxnet.gluon.probability.distributions.exp_family¶

Exponential family class

Classes

ExponentialFamily([event_dim, validate_args])

ExponentialFamily inherits from Distribution. ExponentialFamily is a base class for distributions whose density function has the form: p_F(x;theta) = exp( <t(x), theta> - F(theta) + k(x) ) where t(x): sufficient statistics theta: natural parameters F(theta): log_normalizer k(x): carrier measure.

class mxnet.gluon.probability.distributions.exp_family.ExponentialFamily(event_dim=None, validate_args=None)[source]¶

Bases: Distribution

ExponentialFamily inherits from Distribution. ExponentialFamily is a base class for distributions whose density function has the form: p_F(x;theta) = exp(

<t(x), theta> - F(theta) + k(x)

) where t(x): sufficient statistics theta: natural parameters F(theta): log_normalizer k(x): carrier measure

entropy()[source]¶: Return the entropy of a distribution. The entropy of distributions in exponential families could be computed by: H(P) = F(theta) - <theta, F(theta)’> - E_p[k(x)]