edu::tum::cs::bayesnets::inference::SampleSearch Class Reference

Inheritance diagram for edu::tum::cs::bayesnets::inference::SampleSearch:

Classes
class	SampledAssignment
Public Member Functions
SampledDistribution	_infer () throws Exception
WeightedSample	getWeightedSample (WeightedSample s, int[] nodeOrder, int[] evidenceDomainIndices) throws Exception
	SampleSearch (BeliefNetworkEx bn) throws Exception
Protected Member Functions
int[]	computeNodeOrdering () throws Exception
void	info (int step)
SampledAssignment	sampleForward (BeliefNode node, int[] nodeDomainIndices, boolean[] excluded)
Protected Attributes
int	currentStep
int[]	nodeOrder

Detailed Description

simple implementation of the SampleSearch algorithm by Gogate & Dechter. NOTE: This implementation does not properly weight samples

Definition at line 15 of file SampleSearch.java.

edu::tum::cs::bayesnets::inference::SampleSearch::SampleSearch ( BeliefNetworkEx bn ) throws Exception [inline]

Definition at line 19 of file SampleSearch.java.

SampledDistribution edu::tum::cs::bayesnets::inference::SampleSearch::_infer ( ) throws Exception [inline, virtual]

Definition at line 33 of file SampleSearch.java.

int [] edu::tum::cs::bayesnets::inference::SampleSearch::computeNodeOrdering ( ) throws Exception [inline, protected]

Definition at line 25 of file SampleSearch.java.

WeightedSample edu::tum::cs::bayesnets::inference::SampleSearch::getWeightedSample	(	WeightedSample	s,
		int[]	nodeOrder,
		int[]	evidenceDomainIndices
	)			throws Exception `[inline]`

Definition at line 74 of file SampleSearch.java.

void edu::tum::cs::bayesnets::inference::SampleSearch::info ( int step ) [inline, protected]

Definition at line 29 of file SampleSearch.java.

SampledAssignment edu::tum::cs::bayesnets::inference::SampleSearch::sampleForward	(	BeliefNode	node,
		int[]	nodeDomainIndices,
		boolean[]	excluded
	)			`[inline, protected]`

samples forward, i.e. samples a value for 'node' given its parents

Parameters:

	node	the node for which to sample a value
	nodeDomainIndices	array of domain indices for all nodes in the network; the values for the parents of 'node' must be set already

Returns:: the index of the domain element of 'node' that is sampled, or -1 if sampling is impossible because all entries in the relevant column are 0 TODO should use loopy bp/ijgp to initialize importance distributions rather than sampling from prior

Definition at line 168 of file SampleSearch.java.

Definition at line 17 of file SampleSearch.java.

Definition at line 16 of file SampleSearch.java.

The documentation for this class was generated from the following file: