Module es.upm.fi.cig.multictbnc

Package es.upm.fi.cig.multictbnc.util

Class ProbabilityUtil

java.lang.Object

es.upm.fi.cig.multictbnc.util.ProbabilityUtil

public final class ProbabilityUtil
extends Object

Utility class with methods related to the estimation of probabilities.

Method Summary

Modifier and Type	Method	Description
static double	extremeProbability()	Returns a probability between 0 or 0.3, or between 0.7 and 1.
static <NodeTypeCTBN> double	logLikelihoodSequence(Sequence sequence, List<NodeTypeCTBN> nodesCTBN, State stateClassVariables)	Computes the log-likelihood of a sequence, also known as temporal likelihood (Stella and Amer 2012), given the state of the class variables.
static <NodeTypeBN extends Node> double	logPriorProbabilityClassVariables(List<NodeTypeBN> nodesBN, State stateCVs)	Computes the logarithm of the prior probability of the class variables taking certain values.
static double	marginalLogLikelihoodSequence(double[] laps)	Computes the marginal log-likelihood of a sequence given the unnormalised log-a-posteriori probability for each class configuration.

Methods inherited from class java.lang.Object

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

Method Details

extremeProbability

public static double extremeProbability()

Returns a probability between 0 or 0.3, or between 0.7 and 1.

Returns:

a probability

logLikelihoodSequence

public static <NodeTypeCTBN> double logLikelihoodSequence(Sequence sequence,
 List<NodeTypeCTBN> nodesCTBN,
 State stateClassVariables)

Computes the log-likelihood of a sequence, also known as temporal likelihood (Stella and Amer 2012), given the state of the class variables. This is done by using the CTBN.

Type Parameters:

NodeTypeCTBN - type of the nodes of the continuous-time Bayesian network

Parameters:

sequence - sequence evaluated

nodesCTBN - nodes of a continuous-time Bayesian network

stateClassVariables - class configuration

Returns:

log-likelihood of the sequence given the class configuration

logPriorProbabilityClassVariables

public static <NodeTypeBN extends Node> double logPriorProbabilityClassVariables(List<NodeTypeBN> nodesBN,
 State stateCVs)

Computes the logarithm of the prior probability of the class variables taking certain values. Their probability is computed by using the Bayesian network.

Type Parameters:

NodeTypeBN - type of the nodes of the Bayesian network

Parameters:

nodesBN - nodes of a Bayesian network

stateCVs - class configuration

Returns:

logarithm of the prior probability of the class variables

marginalLogLikelihoodSequence

public static double marginalLogLikelihoodSequence(double[] laps)

Computes the marginal log-likelihood of a sequence given the unnormalised log-a-posteriori probability for each class configuration.

Parameters:

laps - unnormalised log-a-posteriori probabilities

Returns:

marginal log-likelihood of a sequence