Package edu.cmu.tetrad.bayes

Class ApproximateUpdater

java.lang.Object

edu.cmu.tetrad.bayes.ApproximateUpdater

All Implemented Interfaces:

BayesUpdater, ManipulatingBayesUpdater, TetradSerializable, Serializable

public final class ApproximateUpdater
extends Object
implements ManipulatingBayesUpdater

Calculates updated marginals for a Bayes net by simulating data and calculating likelihood ratios. The method is as follows. For P(A | B), enough sample points are simulated from the underlying BayesIm so that 1000 satisfy the condition B. Then the maximum likelihood estimate of condition A is calculated.

Version:

$Id: $Id

Author:

josephramsey

See Also:

• Serialized Form

Constructor Summary

Constructors

Constructor	Description
ApproximateUpdater(BayesIm bayesIm)	Constructs a new updater for the given Bayes net.
ApproximateUpdater(BayesIm bayesIm, Evidence evidence)	Constructs a new updater for the given Bayes net.

Method Summary

Modifier and Type	Method	Description
double[]	calculatePriorMarginals(int nodeIndex)	Calculates the prior marginal probabilities of the given node.
double[]	calculateUpdatedMarginals(int nodeIndex)	Calculates the updated marginal probabilities of the given node, given the evidence.
BayesIm	getBayesIm()	Getter for the field bayesIm.
Evidence	getEvidence()	Getter for the field evidence.
double	getJointMarginal(int[] variables, int[] values)	Computes the joint marginal probability for the specified variables and their corresponding values.
BayesIm	getManipulatedBayesIm()	Getter for the field manipulatedBayesIm.
Graph	getManipulatedGraph()	getManipulatedGraph.
double	getMarginal(int variable, int value)	Returns the marginal probability of the given variable taking the given value, given the evidence.
BayesIm	getUpdatedBayesIm()	getUpdatedBayesIm.
boolean	isJointMarginalSupported()	isJointMarginalSupported.
static ApproximateUpdater	serializableInstance()	Returns a simple exemplar of this class to test serialization.
void	setEvidence(Evidence evidence)	Sets new evidence for the updater.
String	toString()	Prints out the most recent marginal.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Details

ApproximateUpdater

public ApproximateUpdater(BayesIm bayesIm)

Constructs a new updater for the given Bayes net.

Parameters:

bayesIm - the Bayes net to be updated.

ApproximateUpdater

public ApproximateUpdater(BayesIm bayesIm,
 Evidence evidence)

Constructs a new updater for the given Bayes net.

Parameters:

bayesIm - the Bayes net to be updated.

evidence - the evidence for the update.

Method Details

serializableInstance

public static ApproximateUpdater serializableInstance()

Returns a simple exemplar of this class to test serialization.

Returns:

a simple exemplar of this class to test serialization.

getBayesIm

public BayesIm getBayesIm()

Getter for the field bayesIm.

Specified by:

getBayesIm in interface BayesUpdater

Returns:

the Bayes instantiated model that is being updated.

getManipulatedBayesIm

public BayesIm getManipulatedBayesIm()

Getter for the field manipulatedBayesIm.

Specified by:

getManipulatedBayesIm in interface ManipulatingBayesUpdater

Returns:

the Bayes instantiated model after manipulations have been applied.

getManipulatedGraph

public Graph getManipulatedGraph()

getManipulatedGraph.

Specified by:

getManipulatedGraph in interface ManipulatingBayesUpdater

Returns:

the graph for getManipulatedBayesIm().

getUpdatedBayesIm

public BayesIm getUpdatedBayesIm()

getUpdatedBayesIm.

Specified by:

getUpdatedBayesIm in interface ManipulatingBayesUpdater

Returns:

the updated Bayes IM, or null if there is no updated Bayes IM.

getEvidence

public Evidence getEvidence()

Getter for the field evidence.

Specified by:

getEvidence in interface ManipulatingBayesUpdater

Returns:

a copy of the getModel evidence.

setEvidence

public void setEvidence(Evidence evidence)

Sets new evidence for the updater. Once this is called, old updating results should not longer be available.

Sets new evidence for the updater. Once this is called, old updating results should not longer be available.

Sets new evidence for the next update operation.

Specified by:

setEvidence in interface BayesUpdater

Specified by:

setEvidence in interface ManipulatingBayesUpdater

Parameters:

evidence - evidence

getMarginal

public double getMarginal(int variable,
 int value)

Returns the marginal probability of the given variable taking the given value, given the evidence.

Returns the updated graph. This is the graph in which all probabilities of variables conditional on their parents have been updated.

Specified by:

getMarginal in interface BayesUpdater

Specified by:

getMarginal in interface ManipulatingBayesUpdater

Parameters:

variable - variable index

value - category index

Returns:

P(variable = value | evidence), where evidence is getEvidence().

isJointMarginalSupported

public boolean isJointMarginalSupported()

isJointMarginalSupported.

Specified by:

isJointMarginalSupported in interface BayesUpdater

Returns:

a boolean

getJointMarginal

public double getJointMarginal(int[] variables,
 int[] values)

Computes the joint marginal probability for the specified variables and their corresponding values.

Specified by:

getJointMarginal in interface BayesUpdater

Parameters:

variables - an array of integers representing the indices of the variables for which the joint marginal is to be computed.

values - an array of integers representing the corresponding values of the variables for which the joint marginal is to be computed.

Returns:

the joint marginal probability for the specified variables and values.

calculatePriorMarginals

public double[] calculatePriorMarginals(int nodeIndex)

Calculates the prior marginal probabilities of the given node.

Specified by:

calculatePriorMarginals in interface BayesUpdater

Parameters:

nodeIndex - node index

Returns:

P(node = value), where value is the value of the node in the Bayes net.

calculateUpdatedMarginals

public double[] calculateUpdatedMarginals(int nodeIndex)

Calculates the updated marginal probabilities of the given node, given the evidence.

Specified by:

calculateUpdatedMarginals in interface BayesUpdater

Parameters:

nodeIndex - node index

Returns:

P(node = value | evidence), where value is the value of the node in the Bayes net.

toString

public String toString()

Prints out the most recent marginal.

Overrides:

toString in class Object

Returns:

a String object