diva.sketch.classification

Class RubineClassifier

java.lang.Object

diva.sketch.classification.RubineClassifier

All Implemented Interfaces:

Classifier, TrainableClassifier

public class RubineClassifier
extends Object
implements TrainableClassifier

This classifier implements the classic linear discriminator. More information can be obtained from Rubine's paper.

Version:

$Revision: 1.5 $

Author:

Heloise Hse (hwawen@eecs.berkeley.edu)

Constructor Summary

Constructors

Constructor and Description
RubineClassifier()

Method Summary

Modifier and Type	Method and Description
Classification	classify(FeatureSet fs) Return a classification for the specified feature set, or throw a runtime exception if the feature set that is given does not have the same number of features as the training set that it was trained on.
Classification	classify2(FeatureSet fs) Return a classification for the specified feature set, or throw a runtime exception if the feature set that is given does not have the same number of features as the training set that it was trained on.
void	clear() Clear all results of previous trainings (presumably so that this classifier can be trained again from scratch).
boolean	isIncremental() Return whether this classifier is incremental, i.e.
static void	main(String[] argv)
static double	myInvert(double[][] ym, double[][] rm) A new invert routine taken directly from Amulet's gest_matrix.cc (where it's called InvertMatrix).
static double	quadraticForm(double[] v, double[][] m)
void	train(TrainingSet tset) Train the classifier with a given training set and produce a weight vector per training class (a weight for each feature, _weights[]).

Methods inherited from class java.lang.Object

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

Constructor Detail

RubineClassifier

public RubineClassifier()

Method Detail

main

public static void main(String[] argv)

classify2

public Classification classify2(FeatureSet fs)
                         throws ClassifierException

Return a classification for the specified feature set, or throw a runtime exception if the feature set that is given does not have the same number of features as the training set that it was trained on.

Throws:

ClassifierException

classify

public Classification classify(FeatureSet fs)
                        throws ClassifierException

Description copied from interface: Classifier

Return a classification for the specified feature set, or throw a runtime exception if the feature set that is given does not have the same number of features as the training set that it was trained on.

Specified by:

classify in interface Classifier

Throws:

ClassifierException

clear

public void clear()

Clear all results of previous trainings (presumably so that this classifier can be trained again from scratch).

Specified by:

clear in interface TrainableClassifier

isIncremental

public boolean isIncremental()

Return whether this classifier is incremental, i.e. whether this classifier can support multiple calls to "train".

Specified by:

isIncremental in interface TrainableClassifier

train

public void train(TrainingSet tset)
           throws ClassifierException

Train the classifier with a given training set and produce a weight vector per training class (a weight for each feature, _weights[]). It is assumed that all training classes use the same set of features. Also a normalizing constant for each class is computed as well (_constants[]). This method throws a ClassifierException if the training set is not self consisistent, i.e. the feature sets that it contains do not have the same number of features in them. An if matrix inversion fails due to singular matrix.

Specified by:

train in interface TrainableClassifier

Throws:

ClassifierException

myInvert

public static double myInvert(double[][] ym,
                              double[][] rm)

A new invert routine taken directly from Amulet's gest_matrix.cc (where it's called InvertMatrix). Their comment is Matrix inversion using full pivoting. The standard Gauss-Jordan method is used. The return value is the determinant. The input matrix may be the same as the result matrix det = InvertMatrix(inputmatrix, resultmatrix); HISTORY 26-Feb-82 David Smith (drs) at Carnegie-Mellon University Written. Sun Mar 20 19:36:16 EST 1988 - stolen by dandb, and converted to this form

quadraticForm

public static double quadraticForm(double[] v,
                                   double[][] m)