Module: tfp.experimental.vi

Experimental methods and objectives for variational inference.

Modules

util module: Experimental methods and objectives for variational inference.

Functions

build_affine_surrogate_posterior(...): Builds a joint variational posterior with a given event_shape.

build_affine_surrogate_posterior_from_base_distribution(...): Builds a variational posterior by linearly transforming base distributions.

build_affine_surrogate_posterior_from_base_distribution_stateless(...): Builds a variational posterior by linearly transforming base distributions.

build_affine_surrogate_posterior_stateless(...): Builds a joint variational posterior with a given event_shape.

build_asvi_surrogate_posterior(...): Builds a structured surrogate posterior inspired by conjugate updating.

build_asvi_surrogate_posterior_stateless(...): Builds a structured surrogate posterior inspired by conjugate updating.

build_factored_surrogate_posterior(...): Builds a joint variational posterior that factors over model variables.

build_factored_surrogate_posterior_stateless(...): Builds a joint variational posterior that factors over model variables.

build_split_flow_surrogate_posterior(...): Builds a joint variational posterior by splitting a normalizing flow.

Other Members
ASVI_DEFAULT_PRIOR_SUBSTITUTION_RULES	((<class 'tensorflow_probability.python.distributions.half_normal.HalfNormal'>, <function <lambda>>), (<class 'tensorflow_probability.python.distributions.uniform.Uniform'>, <function <lambda>>), (<class 'tensorflow_probability.python.distributions.exponential.Exponential'>, <function <lambda>>), (<class 'tensorflow_probability.python.distributions.chi2.Chi2'>, <function <lambda>>))
ASVI_DEFAULT_SURROGATE_RULES	((<function <lambda>>, <function _asvi_surrogate_rule.<locals>.wrap.<locals>.<lambda>>), (<class 'tensorflow_probability.python.distributions.sample.Sample'>, <function _asvi_surrogate_for_sample>), (<class 'tensorflow_probability.python.distributions.independent.Independent'>, <function _asvi_surrogate_rule.<locals>.wrap.<locals>.<lambda>>), (<function <lambda>>, <function _asvi_surrogate_rule.<locals>.wrap.<locals>.<lambda>>), (<class 'tensorflow_probability.python.distributions.markov_chain.MarkovChain'>, <function _asvi_surrogate_rule.<locals>.wrap.<locals>.<lambda>>))