Package cc.mallet.classify

Class RankMaxEnt

java.lang.Object

cc.mallet.classify.Classifier

cc.mallet.classify.MaxEnt

cc.mallet.classify.RankMaxEnt

All Implemented Interfaces:

AlphabetCarrying, java.io.Serializable

public class RankMaxEnt
extends MaxEnt

Rank Maximum Entropy classifier. This classifier chooses among a set of Instances with binary labels. Expects Instance data to be a FeatureVectorSequence, and the target to be a String representation of the index of the true best FeatureVectorSequence. Note that the Instance target may be a Labels to indicate a tie for the best Instance.

Author:

Aron Culotta culotta@cs.umass.edu

See Also:

Serialized Form

Field Summary

Fields inherited from class cc.mallet.classify.MaxEnt

defaultFeatureIndex, featureSelection, parameters, perClassFeatureSelection

Fields inherited from class cc.mallet.classify.Classifier

instancePipe

Constructor Summary

Constructors

Constructor	Description
RankMaxEnt(Pipe dataPipe, double[] parameters)
RankMaxEnt(Pipe dataPipe, double[] parameters, FeatureSelection featureSelection)
RankMaxEnt(Pipe dataPipe, double[] parameters, FeatureSelection[] perClassFeatureSelection)
RankMaxEnt(Pipe dataPipe, double[] parameters, FeatureSelection featureSelection, FeatureSelection[] perClassFeatureSelection)

Method Summary

Modifier and Type	Method	Description
Classification	classify(Instance instance)
void	getClassificationScores(Instance instance, double[] scores)
void	getClassificationScoresForTies(Instance instance, double[] scores, int[] bestLabels)	Used by RankMaxEntTrainer to calculate the value when the labeling contains ties.
void	getUnnormalizedClassificationScores(Instance instance, double[] scores)	returns unnormalized scores, corresponding to the score an element of the InstanceList being the "top" instance
void	print()	Outputs human-readable description of classifier (e.g., list of weights, decision tree) to System.out

Methods inherited from class cc.mallet.classify.MaxEnt

getClassificationScoresWithTemperature, getDefaultFeatureIndex, getFeatureSelection, getNumParameters, getNumParameters, getParameters, getPerClassFeatureSelection, print, print, printExtremeFeatures, printRank, setDefaultFeatureIndex, setFeatureSelection, setParameter, setParameters, setPerClassFeatureSelection

Methods inherited from class cc.mallet.classify.Classifier

alphabetsMatch, classify, classify, classify, getAccuracy, getAlphabet, getAlphabets, getAverageRank, getF1, getF1, getF1, getInstancePipe, getLabelAlphabet, getPrecision, getPrecision, getPrecision, getRecall, getRecall, getRecall

Methods inherited from class java.lang.Object

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

Constructor Detail

RankMaxEnt

public RankMaxEnt(Pipe dataPipe,
                  double[] parameters,
                  FeatureSelection featureSelection,
                  FeatureSelection[] perClassFeatureSelection)

RankMaxEnt

public RankMaxEnt(Pipe dataPipe,
                  double[] parameters,
                  FeatureSelection featureSelection)

RankMaxEnt

public RankMaxEnt(Pipe dataPipe,
                  double[] parameters,
                  FeatureSelection[] perClassFeatureSelection)

RankMaxEnt

public RankMaxEnt(Pipe dataPipe,
                  double[] parameters)

Method Detail

getUnnormalizedClassificationScores

public void getUnnormalizedClassificationScores(Instance instance,
                                                double[] scores)

returns unnormalized scores, corresponding to the score an element of the InstanceList being the "top" instance

Overrides:

getUnnormalizedClassificationScores in class MaxEnt

Parameters:

instance - instance with data field a InstanceList.

scores - has length = number of Instances in Instance.data, which is of type InstanceList

getClassificationScores

public void getClassificationScores(Instance instance,
                                    double[] scores)

Overrides:

getClassificationScores in class MaxEnt

getClassificationScoresForTies

public void getClassificationScoresForTies(Instance instance,
                                           double[] scores,
                                           int[] bestLabels)

Used by RankMaxEntTrainer to calculate the value when the labeling contains ties. Does not include scores of tied elements in normalization.

Parameters:

instance -

scores -

bestLabels - Indices of Instances ties for 1st place.

classify

public Classification classify(Instance instance)

Overrides:

classify in class MaxEnt

print

public void print()

Description copied from class: Classifier

Outputs human-readable description of classifier (e.g., list of weights, decision tree) to System.out

Overrides:

print in class MaxEnt