io.improbable.keanu.algorithms

Interface ProbabilisticModelWithGradient

All Superinterfaces:

java.lang.AutoCloseable, ProbabilisticModel

All Known Implementing Classes:

KeanuProbabilisticModelWithGradient

public interface ProbabilisticModelWithGradient
extends ProbabilisticModel

A ProbabilisticModel that can also return the gradient of its log-probability: required by algorithms such as Keanu.Sampling.NUTS and GradientOptimizer.

Method Summary

Modifier and Type	Method and Description
java.util.Map<VariableReference,DoubleTensor>	logLikelihoodGradients()
java.util.Map<VariableReference,DoubleTensor>	logLikelihoodGradients(java.util.Map<VariableReference,?> inputs)
java.util.Map<VariableReference,DoubleTensor>	logProbGradients()
java.util.Map<VariableReference,DoubleTensor>	logProbGradients(java.util.Map<VariableReference,?> inputs)

Methods inherited from interface io.improbable.keanu.algorithms.ProbabilisticModel

close, getContinuousLatentVariables, getLatentVariables, logLikelihood, logLikelihood, logProb, logProb, logProbAfter

Method Detail

logProbGradients

java.util.Map<VariableReference,DoubleTensor> logProbGradients(java.util.Map<VariableReference,?> inputs)

logProbGradients

java.util.Map<VariableReference,DoubleTensor> logProbGradients()

logLikelihoodGradients

java.util.Map<VariableReference,DoubleTensor> logLikelihoodGradients(java.util.Map<VariableReference,?> inputs)

logLikelihoodGradients

java.util.Map<VariableReference,DoubleTensor> logLikelihoodGradients()