io.improbable.keanu.model

Class SamplingModelFitter

java.lang.Object

io.improbable.keanu.model.SamplingModelFitter

All Implemented Interfaces:

ModelFitter

public class SamplingModelFitter
extends java.lang.Object
implements ModelFitter

Constructor Summary

Constructors

Constructor and Description
SamplingModelFitter(java.util.function.Function<KeanuProbabilisticModel,PosteriorSamplingAlgorithm> samplingAlgorithmGenerator, int sampleCount) This fitter uses a PosteriorSamplingAlgorithm, in contrast to the MAPModelFitter and MaximumLikelihoodModelFitter, which use gradient methods.

Method Summary

Modifier and Type	Method and Description
void	fit(ModelGraph modelGraph) Uses a posterior sampling algorithm (e.g.
NetworkSamples	getNetworkSamples()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SamplingModelFitter

public SamplingModelFitter(java.util.function.Function<KeanuProbabilisticModel,PosteriorSamplingAlgorithm> samplingAlgorithmGenerator,
                           int sampleCount)

This fitter uses a PosteriorSamplingAlgorithm, in contrast to the MAPModelFitter and MaximumLikelihoodModelFitter, which use gradient methods. The model's latent vertices will have their values set to the average over the samples.

Parameters:

samplingAlgorithmGenerator - The algorithm to use, e.g. MetropolisHastings

sampleCount - The number of sample points to take.

Method Detail

fit

public void fit(ModelGraph modelGraph)

Uses a posterior sampling algorithm (e.g. Metropolis Hastings) to fit the model graph to the input and output data. This will mutate the graph which can then be used to construct a graph-backed model like, for instance, a RegressionModel

Specified by:

fit in interface ModelFitter

getNetworkSamples

public NetworkSamples getNetworkSamples()