com.mockturtlesolutions.snifflib.invprobs

Class AbstractProposalDistribution

java.lang.Object

com.mockturtlesolutions.snifflib.functions.AbstractFunction

com.mockturtlesolutions.snifflib.invprobs.StatisticalModel

com.mockturtlesolutions.snifflib.invprobs.AbstractProposalDistribution

All Implemented Interfaces:

Nameable, FigureReportable, Reportable, DeclaredParameters, DeclaredVariables, Likelihood, NamedParameters, ObjectSampler, ObjectSamplerClient, OptimizableScalar, ProposalDistribution

Direct Known Subclasses:

GaussianProposal

public abstract class AbstractProposalDistribution
extends StatisticalModel
implements ProposalDistribution

A standard approach to the proposal distribution called for in Metropolis-Hastings Markov Chain Monte-Carlo. Extensions of this class must override the negLogPrior() method from StatisticalModel. The class guarantees that when negLogPrior is called the output of getParams() and getPriorParams() will be appropriately set to the old or new parameters. Note that typically external classes will not call negLogPrior() directly but will rather call negLogLikeOldGivenNew() or negLogLikeNewGivenOld() which are provided through the ProposalDistribution interface. This classes implementation of negLogLikeOldGivenNew() sets the model parameters to the old values and sets the prior parameters to the new values before calling negLogPrior(). Similarly negLogLikeNewGivenOld() sets the model parameters to the new values and sets the prior parameters to the old values before calling negLogPrior(). Thus, a typical user implemented extension to this class need only over-ride negLogPrior()

Field Summary

Fields

Modifier and Type	Field and Description
protected DblParamSet	mixingParameter
protected DblParamSet	newParams
protected DblParamSet	oldParams

Fields inherited from class com.mockturtlesolutions.snifflib.invprobs.StatisticalModel

adapter, FigureReporters, invertParameterDeclarations, LIKELIHOOD, NEG_LOG_LIKELIHOOD, paramListeners, Params, PosteriorParams, PREDICTION, priorParamListeners, PriorParams, SUM_NEG_LOG_LIKELIHOOD, usingDataSet

Fields inherited from class com.mockturtlesolutions.snifflib.functions.AbstractFunction

domainDimension, rangeDimension

Constructor Summary

Constructors

Constructor and Description
AbstractProposalDistribution(StatisticalModel model) The means of the Gaussians will be assumed to be the current values in

Method Summary

Methods

Modifier and Type	Method and Description
DblMatrix	getMixingParam(java.lang.String name)
DblParamSet	getMixingParams()
DblParamSet	negLogLikeNewGivenOld()
DblParamSet	negLogLikeOldGivenNew()
abstract DblParamSet	negLogPrior() Returns the negative log of the prior probability of the current parameter set.
void	replaceMixingParams(DblParamSet M)
DblParamSet	sample()
DblParamSet	sample(java.lang.String[] subset)
DblParamSet	samplePosterior() Returns a sample from the posterior distribution of the parameters.
abstract DblParamSet	samplePosterior(java.lang.String[] subset) Returns a sample from the posterior distribution of the given subset of parameters.
void	setMixingParam(java.lang.String name, DblMatrix x)
void	setMixingParams(DblParamSet M)
void	updateNewParams(DblParamSet params)
void	updateOldParams(DblParamSet params)

Methods inherited from class com.mockturtlesolutions.snifflib.invprobs.StatisticalModel

about, addFigureReporter, addObjectToSample, addParameterChangeListener, declareIfNotParameter, declareIfNotPriorParameter, declareParameter, declareParameter, declarePriorParameter, declarePriorParameter, declareVariable, fireParameterAdded, fireParameterRemoved, fireParameterValueChanged, fireParameterValuesChanged, firePriorParameterAdded, firePriorParameterRemoved, firePriorParameterValueChanged, firePriorParameterValuesChanged, getDataAugment, getDataSet, getDataSetAdapter, getDeclaredParameters, getDeclaredPriorParameters, getDeclaredVariables, getFigureReporter, getFigureReporters, getIgnoreNaN, getMSE, getName, getNewObjectSample, getNewObjectSamples, getObjectSamplingOffered, getObjectsToSample, getParam, getParams, getPosteriorParam, getPosteriorParams, getPredictionAt, getPriorParam, getPriorParams, getValueAt, getValueAt, getValueAt, getValueToMinimize, getX, getY, hasParameter, hasPriorParameter, invertParameterDeclarations, isSamplingObject, isVectorBased, likelihood, mapName, negLogLikelihood, numParams, offerObjectSamplingFor, offersObjectSamplingFor, parameterSet, putNewObjectSample, removeAllParameterChangeListener, removeFigureReporter, removeObjectToSample, removeParameterChangeListener, replaceParams, replacePosteriorParams, replacePriorParams, residuals, revokeObjectSamplingFor, revokeParameter, revokePriorParameter, revokeVariable, setDataAugment, setDataSet, setIgnoreNaN, setName, setObjectsToSample, setParam, setParams, setPosteriorParam, setPosteriorParams, setPriorParam, setPriorParams, setValueToGet, setX, setY, sumNegLogLikelihood, unMapName, variableSet

Methods inherited from class com.mockturtlesolutions.snifflib.functions.AbstractFunction

domainDimension, plotOver, plotOver, rangeDimension

Methods inherited from class java.lang.Object

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

Field Detail

mixingParameter

protected DblParamSet mixingParameter

newParams

protected DblParamSet newParams

oldParams

protected DblParamSet oldParams

Constructor Detail

AbstractProposalDistribution

public AbstractProposalDistribution(StatisticalModel model)

The means of the Gaussians will be assumed to be the current values in

Method Detail

sample

public DblParamSet sample(java.lang.String[] subset)

Specified by:

sample in interface ProposalDistribution

sample

public DblParamSet sample()

Specified by:

sample in interface ProposalDistribution

updateNewParams

public void updateNewParams(DblParamSet params)

Specified by:

updateNewParams in interface ProposalDistribution

updateOldParams

public void updateOldParams(DblParamSet params)

Specified by:

updateOldParams in interface ProposalDistribution

samplePosterior

public DblParamSet samplePosterior()

Description copied from class: StatisticalModel

Returns a sample from the posterior distribution of the parameters. This default implementation returns the current parameter set.

Overrides:

samplePosterior in class StatisticalModel

samplePosterior

public abstract DblParamSet samplePosterior(java.lang.String[] subset)

Description copied from class: StatisticalModel

Returns a sample from the posterior distribution of the given subset of parameters. Subset of parameters are given as a Vector of Strings. This default implementation returns the current parameter set.

Overrides:

samplePosterior in class StatisticalModel

negLogPrior

public abstract DblParamSet negLogPrior()

Description copied from class: StatisticalModel

Returns the negative log of the prior probability of the current parameter set. This default implementation assumes a prior probability of 1 for each parameter resulting in a return value of zero. Method should be over-riden for other priors.

Overrides:

negLogPrior in class StatisticalModel

negLogLikeOldGivenNew

public DblParamSet negLogLikeOldGivenNew()

Specified by:

negLogLikeOldGivenNew in interface ProposalDistribution

negLogLikeNewGivenOld

public DblParamSet negLogLikeNewGivenOld()

Specified by:

negLogLikeNewGivenOld in interface ProposalDistribution

getMixingParams

public DblParamSet getMixingParams()

Specified by:

getMixingParams in interface ProposalDistribution

setMixingParam

public void setMixingParam(java.lang.String name,
                  DblMatrix x)

Specified by:

setMixingParam in interface ProposalDistribution

getMixingParam

public DblMatrix getMixingParam(java.lang.String name)

Specified by:

getMixingParam in interface ProposalDistribution

setMixingParams

public void setMixingParams(DblParamSet M)

Specified by:

setMixingParams in interface ProposalDistribution

replaceMixingParams

public void replaceMixingParams(DblParamSet M)

Specified by:

replaceMixingParams in interface ProposalDistribution