jsat.classifiers.bayesian.graphicalmodel

Class K2NetworkLearner

java.lang.Object

jsat.classifiers.bayesian.graphicalmodel.DiscreteBayesNetwork

jsat.classifiers.bayesian.graphicalmodel.K2NetworkLearner

All Implemented Interfaces:

Serializable, Cloneable, Classifier

public class K2NetworkLearner
extends DiscreteBayesNetwork

An implementation of the K2 algorithm for learning the structure of a Bayesian Network. When trained, if no network has been specified, the K2 algorithm will attempt to learn a network structure. The network structure can also be learned by calling learnNetwork(jsat.classifiers.ClassificationDataSet) directly.

Note, that the K2 algorithm attempts to learn a whole network structure, and may learn things that are relevant for the classification task. K2 often does not provide satisfactory results for classification.

See: A bayesian method for the induction of probabilistic networks from data. Gregory F. Cooper and Edward Herskovits.

Author:

Edward Raff

See Also:

Serialized Form

Field Summary

Fields inherited from class jsat.classifiers.bayesian.graphicalmodel.DiscreteBayesNetwork

cpts, dag, DEFAULT_USE_PRIORS, predicting, priors

Constructor Summary

Constructors

Constructor and Description
K2NetworkLearner()

Method Summary

Modifier and Type	Method and Description
double	f(int i, Set<Integer> pi, ClassificationDataSet D)
int	getMaxParents() Returns the maximum number of parents allowed when learning a network structure, or zero if any number of parents are valid.
void	learnNetwork(ClassificationDataSet D) Learns the network structure from the given data set.
void	setMaxParents(int maxParents) Sets the maximum number of parents to allow a node when learning the network structure.
void	trainC(ClassificationDataSet dataSet) Trains the classifier and constructs a model for classification using the given data set.

Methods inherited from class jsat.classifiers.bayesian.graphicalmodel.DiscreteBayesNetwork

classify, clone, depends, supportsWeightedData, trainC

Methods inherited from class java.lang.Object

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

Constructor Detail

K2NetworkLearner

public K2NetworkLearner()

Method Detail

setMaxParents

public void setMaxParents(int maxParents)

Sets the maximum number of parents to allow a node when learning the network structure. If a non zero value is supplied, nodes will be allowed any number of parents.

Parameters:

maxParents - sets the maximum number of parents a node may learn

getMaxParents

public int getMaxParents()

Returns the maximum number of parents allowed when learning a network structure, or zero if any number of parents are valid.

Returns:

the maximum number of parents a node man learn

learnNetwork

public void learnNetwork(ClassificationDataSet D)

Learns the network structure from the given data set.

Parameters:

D - the data set to learn the network from

trainC

public void trainC(ClassificationDataSet dataSet)

Description copied from interface: Classifier

Trains the classifier and constructs a model for classification using the given data set.

Specified by:

trainC in interface Classifier

Overrides:

trainC in class DiscreteBayesNetwork

Parameters:

dataSet - the data set to train on

f

public double f(int i,
                Set<Integer> pi,
                ClassificationDataSet D)