cc.mallet.classify.constraints.ge
Class MaxEntFLGEConstraints

java.lang.Object
  cc.mallet.classify.constraints.ge.MaxEntFLGEConstraints

All Implemented Interfaces:

MaxEntGEConstraint

Direct Known Subclasses:

MaxEntKLFLGEConstraints, MaxEntL2FLGEConstraints

public abstract class MaxEntFLGEConstraints
extends java.lang.Object
implements MaxEntGEConstraint

Abstract expectation constraint for use with Generalized Expectation (GE).

Author:

Gregory Druck

Nested Class Summary

protected class

MaxEntFLGEConstraints.MaxEntFLGEConstraint

Field Summary

protected gnu.trove.TIntObjectHashMap<MaxEntFLGEConstraints.MaxEntFLGEConstraint>	constraints
protected gnu.trove.TIntArrayList	indexCache
protected int	numFeatures
protected int	numLabels
protected boolean	useValues
protected gnu.trove.TDoubleArrayList	valueCache

Constructor Summary

MaxEntFLGEConstraints(int numFeatures, int numLabels, boolean useValues)

Method Summary

abstract void	addConstraint(int fi, double[] ex, double weight)
void	computeExpectations(FeatureVector input, double[] dist, double weight) Compute expectations using provided distribution over labels.
double	getCompositeConstraintFeatureValue(FeatureVector input, int label) Computes the composite constraint feature value (over all constraint features) for FeatureVector fv and label label.
void	preProcess(FeatureVector input) Gives the constraint the option to do some caching using only the FeatureVector.
java.util.BitSet	preProcess(InstanceList data)
void	zeroExpectations() Zero expectation values.

Methods inherited from class java.lang.Object

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

Methods inherited from interface cc.mallet.classify.constraints.ge.MaxEntGEConstraint

getValue

Field Detail

useValues

protected boolean useValues

numLabels

protected int numLabels

numFeatures

protected int numFeatures

constraints

protected gnu.trove.TIntObjectHashMap<MaxEntFLGEConstraints.MaxEntFLGEConstraint> constraints

indexCache

protected gnu.trove.TIntArrayList indexCache

valueCache

protected gnu.trove.TDoubleArrayList valueCache

Constructor Detail

MaxEntFLGEConstraints

public MaxEntFLGEConstraints(int numFeatures,
                             int numLabels,
                             boolean useValues)

Method Detail

addConstraint

public abstract void addConstraint(int fi,
                                   double[] ex,
                                   double weight)

getCompositeConstraintFeatureValue

public double getCompositeConstraintFeatureValue(FeatureVector input,
                                                 int label)

Description copied from interface: MaxEntGEConstraint

Computes the composite constraint feature value (over all constraint features) for FeatureVector fv and label label.

Specified by:

getCompositeConstraintFeatureValue in interface MaxEntGEConstraint

Parameters:

input - input FeatureVector

label - output label index

Returns:

Constraint feature value

computeExpectations

public void computeExpectations(FeatureVector input,
                                double[] dist,
                                double weight)

Description copied from interface: MaxEntGEConstraint

Compute expectations using provided distribution over labels.

Specified by:

computeExpectations in interface MaxEntGEConstraint

Parameters:

input - FeatureVector

dist - Distribution over labels

zeroExpectations

public void zeroExpectations()

Description copied from interface: MaxEntGEConstraint

Zero expectation values. Called before re-computing gradient.

Specified by:

zeroExpectations in interface MaxEntGEConstraint

preProcess

public java.util.BitSet preProcess(InstanceList data)

Specified by:

preProcess in interface MaxEntGEConstraint

Parameters:

data - Unlabeled data

Returns:

Returns a bitset of the size of the data, with the bit set if a constraint feature fires in that instance.

preProcess

public void preProcess(FeatureVector input)

Description copied from interface: MaxEntGEConstraint

Gives the constraint the option to do some caching using only the FeatureVector. For example, the constrained input features could be cached.

Specified by:

preProcess in interface MaxEntGEConstraint

Parameters:

input - FeatureVector input