mrftools.ApproxMaxLikelihood module

Class to do generative learning directly on MRF parameters.

class mrftools.ApproxMaxLikelihood.ApproxMaxLikelihood(markov_net, inference_type=<class 'mrftools.MatrixBeliefPropagator.MatrixBeliefPropagator'>)

Bases: mrftools.Learner.Learner

Object that runs approximate maximum likelihood parameter training. This method creates an indicator model where every feature is an indicator function, also known as an overcomplete representation.

add_data(labels)

Add observed training data

Parameters:labels (dict) – dictionary containing an integer state value for each observed variable Returns:None