Package com.bayesserver.learning.parameters

Class ParameterLearningOutput

java.lang.Object

com.bayesserver.learning.parameters.ParameterLearningOutput

All Implemented Interfaces:

Cloneable

public final class ParameterLearningOutput
extends Object
implements Cloneable

Contains summary information returned by ParameterLearning.learn(com.bayesserver.data.EvidenceReaderCommand, com.bayesserver.learning.parameters.ParameterLearningOptions).

Method Summary

Modifier and Type	Method	Description
Double	getBIC()	Gets the Bayesian Information Criterion (BIC) for the final learnt Network based on the learning data.
double	getCaseCount()	Gets the number of cases (records) in the learning data.
boolean	getConverged()	Gets a value indicating whether this parameter learning converged.
int	getIterationCount()	Gets the number of iterations performed during learning.
Double	getLogLikelihood()	Gets the log likelihood of the learning data with the final learnt Network.
int	getSeed()	Gets the seed used to generate random numbers for initialization.
long	getUnweightedCaseCount()	Gets the unweighted case count in the learning data.
double	getWeightedCaseCount()	Gets the weighted case count in the learning data.

Methods inherited from class java.lang.Object

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

Method Detail

getSeed

public int getSeed()

Gets the seed used to generate random numbers for initialization. Only valid when ParameterLearningOptions.getMaximumConcurrency() is 1.

getConverged

public boolean getConverged()

Gets a value indicating whether this parameter learning converged. For more information see ParameterLearningOptions.getTolerance().

Returns:

true if converged; otherwise, false.

getLogLikelihood

public Double getLogLikelihood()

Gets the log likelihood of the learning data with the final learnt Network.

The log likelihood increases as the model fits the data better.

This value is only calculated if ParameterLearningOptions.getCalculateStatistics() is true.

Returns:

The log likelihood.

getIterationCount

public int getIterationCount()

Gets the number of iterations performed during learning.

Returns:

The iteration count.

getCaseCount

public double getCaseCount()

Gets the number of cases (records) in the learning data. Any weights associated with cases are incorporated into this sum.

Returns:

The case count.

getWeightedCaseCount

public double getWeightedCaseCount()

Gets the weighted case count in the learning data. Any weights associated with cases are incorporated into this sum.

Returns:

The weighted case count.

getUnweightedCaseCount

public long getUnweightedCaseCount()

Gets the unweighted case count in the learning data. Any weights associated with cases are NOT incorporated into this sum.

Returns:

The unweighted case count.

getBIC

public Double getBIC()

Gets the Bayesian Information Criterion (BIC) for the final learnt Network based on the learning data.

When comparing two models using the BIC heuristic, a lower value is preferred.

This statistic is only calculated if ParameterLearningOptions.getCalculateStatistics() is true.

The BIC statistics is a Log Likelihood based statistic which penalizes models with more parameters. It can be used as an heuristic to compare networks with different parameters learnt with the same data.

Returns:

The BIC statistic.