edu.cmu.minorthird.classify.algorithms.linear
Class Hyperplane

java.lang.Object
  edu.cmu.minorthird.classify.BinaryClassifier
      edu.cmu.minorthird.classify.algorithms.linear.Hyperplane

All Implemented Interfaces:

Classifier, Visible, java.io.Serializable

public class Hyperplane
extends BinaryClassifier
implements Visible, java.io.Serializable

A weighted combination of features.

Author:

William Cohen

See Also:

Serialized Form

Field Summary

static Feature	BIAS_TERM Weight for an invisible 'bias feature' which is considered to be present in every instance.
protected gnu.trove.TObjectDoubleHashMap	hyperplaneWeights

Fields inherited from class edu.cmu.minorthird.classify.BinaryClassifier

classifierLearner

Constructor Summary

Hyperplane()

Method Summary

java.lang.String	explain(Instance instance) Justify inner product of hyperplane and instance weights.
java.util.Iterator<Feature>	featureIterator() Iterate over all features with non-zero weight.
double	featureScore(Feature feature) Weight for a feature in the hyperplane.
Explanation	getExplanation(Instance instance) Return an Explanation for the classification
boolean	hasFeature(Feature feat) Checks the presence of a feature in hyperplane
void	increment(Feature f, double delta) Increment one feature from the hyperplane by delta
void	increment(Hyperplane b) Add hyperplane b to this hyperplane.
void	increment(Hyperplane b, double delta) Add hyperplane b*delta to this hyperplane.
void	increment(Instance instance, double delta) Add the value of the features in the instance to this hyperplane.
void	incrementBias(double delta) Increment the bias term for the hyperplane by delta
void	multiply(double factor) Multiply all weights by a factor
void	multiply(Feature f, double delta) Multiply one feature from the hyperplane by delta
double	score(Instance instance) Inner product of hyperplane and instance weights.
void	setBias(double delta) Set the bias term for the hyperplane to delta
void	startIgnoringWeights() After this call is made, the hyperplane will assume that all feature weights are one in instances.
Viewer	toGUI() Create a view of this object
java.lang.String	toString()

Methods inherited from class edu.cmu.minorthird.classify.BinaryClassifier

classification, getClassifierLearner, setClassifierLearner

Methods inherited from class java.lang.Object

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

Field Detail

BIAS_TERM

public static final Feature BIAS_TERM

Weight for an invisible 'bias feature' which is considered to be present in every instance. In otherwords, as a classifier, the score of a hyperplane h on an instance x is sum_{feature f} x.score(f)*h.featureScore(f) + h.featureScore(BIAS_TERM).

hyperplaneWeights

protected transient gnu.trove.TObjectDoubleHashMap hyperplaneWeights

Constructor Detail

Hyperplane

public Hyperplane()

Method Detail

startIgnoringWeights

public void startIgnoringWeights()

After this call is made, the hyperplane will assume that all feature weights are one in instances. Specifically in calls to increment(instance,delta) or increment(hyperplane,delta), feature weights will be assumed to be one. For backward compatibility with an old buggy version.

score

public double score(Instance instance)

Inner product of hyperplane and instance weights.

Specified by:

score in class BinaryClassifier

explain

public java.lang.String explain(Instance instance)

Justify inner product of hyperplane and instance weights.

Specified by:

explain in interface Classifier

getExplanation

public Explanation getExplanation(Instance instance)

Description copied from interface: Classifier

Return an Explanation for the classification

Specified by:

getExplanation in interface Classifier

increment

public void increment(Feature f,
                      double delta)

Increment one feature from the hyperplane by delta

incrementBias

public void incrementBias(double delta)

Increment the bias term for the hyperplane by delta

setBias

public void setBias(double delta)

Set the bias term for the hyperplane to delta

increment

public void increment(Instance instance,
                      double delta)

Add the value of the features in the instance to this hyperplane.

multiply

public void multiply(double factor)

Multiply all weights by a factor

multiply

public void multiply(Feature f,
                     double delta)

Multiply one feature from the hyperplane by delta

hasFeature

public boolean hasFeature(Feature feat)

Checks the presence of a feature in hyperplane

increment

public void increment(Hyperplane b,
                      double delta)

Add hyperplane b*delta to this hyperplane.

increment

public void increment(Hyperplane b)

Add hyperplane b to this hyperplane.

featureScore

public double featureScore(Feature feature)

Weight for a feature in the hyperplane.

featureIterator

public java.util.Iterator<Feature> featureIterator()

Iterate over all features with non-zero weight.

toGUI

public Viewer toGUI()

Description copied from interface: Visible

Create a view of this object

Specified by:

toGUI in interface Visible

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object