weka.classifiers.functions
Class SMO.BinarySMO

java.lang.Object
  weka.classifiers.functions.SMO.BinarySMO

All Implemented Interfaces:

java.io.Serializable

Enclosing class:

SMO

private class SMO.BinarySMO

extends java.lang.Object

implements java.io.Serializable

Class for building a binary support vector machine.

Field Summary

private double[]	m_alpha The Lagrange multipliers.
private double	m_b The thresholds.
private double	m_bLow The thresholds.
private double	m_bUp The thresholds.
private double[]	m_class The transformed class values.
private Instances	m_data The training data.
private double[]	m_errors The current set of errors for all non-bound examples.
private SMOset	m_I0 The five different sets used by the algorithm.
private SMOset	m_I1
private SMOset	m_I2
private SMOset	m_I3
private SMOset	m_I4
private int	m_iLow The indices for m_bLow and m_bUp
private int	m_iUp The indices for m_bLow and m_bUp
private Kernel	m_kernel Kernel to use
private Logistic	m_logistic Stores logistic regression model for probability estimate
private int[]	m_sparseIndices
private double[]	m_sparseWeights Variables to hold weight vector in sparse form.
private double	m_sumOfWeights Stores the weight of the training instances
private SMOset	m_supportVectors The set of support vectors
private double[]	m_weights Weight vector for linear machine.

Constructor Summary

private

SMO.BinarySMO()

Method Summary

private void	buildClassifier(Instances insts, int cl1, int cl2, boolean fitLogistic, int numFolds, int randomSeed) Method for building the binary classifier.
private void	checkClassifier() Quick and dirty check whether the quadratic programming problem is solved.
private boolean	examineExample(int i2) Examines instance.
private void	fitLogistic(Instances insts, int cl1, int cl2, int numFolds, java.util.Random random) Fits logistic regression model to SVM outputs analogue to John Platt's method.
private double	SVMOutput(int index, Instance inst) Computes SVM output for given instance.
private boolean	takeStep(int i1, int i2, double F2) Method solving for the Lagrange multipliers for two instances.
java.lang.String	toString() Prints out the classifier.

Methods inherited from class java.lang.Object

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

Field Detail

m_alpha

private double[] m_alpha

The Lagrange multipliers.

m_b

private double m_b

The thresholds.

m_bLow

private double m_bLow

The thresholds.

m_bUp

private double m_bUp

The thresholds.

m_iLow

private int m_iLow

The indices for m_bLow and m_bUp

m_iUp

private int m_iUp

The indices for m_bLow and m_bUp

m_data

private Instances m_data

The training data.

m_weights

private double[] m_weights

Weight vector for linear machine.

m_sparseWeights

private double[] m_sparseWeights

Variables to hold weight vector in sparse form. (To reduce storage requirements.)

m_sparseIndices

private int[] m_sparseIndices

m_kernel

private Kernel m_kernel

Kernel to use

m_class

private double[] m_class

The transformed class values.

m_errors

private double[] m_errors

The current set of errors for all non-bound examples.

m_I0

private SMOset m_I0

The five different sets used by the algorithm.

m_I1

private SMOset m_I1

m_I2

private SMOset m_I2

m_I3

private SMOset m_I3

m_I4

private SMOset m_I4

m_supportVectors

private SMOset m_supportVectors

The set of support vectors

m_logistic

private Logistic m_logistic

Stores logistic regression model for probability estimate

m_sumOfWeights

private double m_sumOfWeights

Stores the weight of the training instances

Constructor Detail

SMO.BinarySMO

private SMO.BinarySMO()

Method Detail

fitLogistic

private void fitLogistic(Instances insts,
                         int cl1,
                         int cl2,
                         int numFolds,
                         java.util.Random random)
                  throws java.lang.Exception

Fits logistic regression model to SVM outputs analogue to John Platt's method.

Parameters:

insts - the set of training instances

cl1 - the first class' index

cl2 - the second class' index

Throws:

java.lang.Exception - if the sigmoid can't be fit successfully

buildClassifier

private void buildClassifier(Instances insts,
                             int cl1,
                             int cl2,
                             boolean fitLogistic,
                             int numFolds,
                             int randomSeed)
                      throws java.lang.Exception

Method for building the binary classifier.

Parameters:

insts - the set of training instances

cl1 - the first class' index

cl2 - the second class' index

fitLogistic - true if logistic model is to be fit

numFolds - number of folds for internal cross-validation

Throws:

java.lang.Exception - if the classifier can't be built successfully

SVMOutput

private double SVMOutput(int index,
                         Instance inst)
                  throws java.lang.Exception

Computes SVM output for given instance.

Parameters:

index - the instance for which output is to be computed

inst - the instance

Returns:

the output of the SVM for the given instance

Throws:

java.lang.Exception

toString

public java.lang.String toString()

Prints out the classifier.

Returns:

a description of the classifier as a string

examineExample

private boolean examineExample(int i2)
                        throws java.lang.Exception

Examines instance.

Parameters:

i2 - index of instance to examine

Returns:

true if examination was successfull

Throws:

java.lang.Exception - if something goes wrong

takeStep

private boolean takeStep(int i1,
                         int i2,
                         double F2)
                  throws java.lang.Exception

Method solving for the Lagrange multipliers for two instances.

Parameters:

i1 - index of the first instance

i2 - index of the second instance

Returns:

true if multipliers could be found

Throws:

java.lang.Exception - if something goes wrong

checkClassifier

private void checkClassifier()
                      throws java.lang.Exception

Quick and dirty check whether the quadratic programming problem is solved.

Throws:

java.lang.Exception