org.knime.base.node.mine.svm
Class Svm

java.lang.Object
  org.knime.base.node.mine.svm.Svm

public class Svm
extends Object

This class represents a (binary) support vector machine. It works by remembering the support vectors and the corresponding alpha values.

Author:

Stefan, University of Konstanz, cebron, University of Konstanz

Constructor Summary

Svm(DoubleVector[] supportVectors, double[] alpha, String positive, double b, Kernel kernel)
Constructor.

Svm(ModelContentRO predParams, String id)
Loads a binary SVM from a ModelContent object.

Method Summary

double	distance(DoubleVector vector) Computes the distance from the hyperplane in the kernel induced hyperspace.
double[]	getAlphas()
double	getMargin() The margin of a SVM is the minimum distance from all support vectors to the decision hyperplane.
String	getPositive()
DoubleVector[]	getSupportVectors()
double[]	getTargetAlphas() Multiplies the alpha coefficients of the SVM with the target value (-1 or 1, depending if the input is a 'positive' example or not).
double	getThreshold()
double	predict(DoubleVector vector) Computes the predicted value of a vector by using the current SVM.
void	saveToPredictorParams(ModelContentWO predParams, String id) Save the Support Vector Machine for later use.
String	toString()

Methods inherited from class java.lang.Object

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

Constructor Detail

Svm

public Svm(DoubleVector[] supportVectors,
           double[] alpha,
           String positive,
           double b,
           Kernel kernel)

Constructor.

Parameters:

supportVectors - the support vectors that define the SVM

alpha - the corresponding Lagrange coefficients

positive - the class for which SVM should yield 1

b - the threshold

kernel - the kernel to use

Svm

public Svm(ModelContentRO predParams,
           String id)
    throws InvalidSettingsException

Loads a binary SVM from a ModelContent object.

Parameters:

predParams - the object to read the SVM configuration from.

id - the unique identifier

Throws:

InvalidSettingsException - if the required keys are not present

Method Detail

distance

public double distance(DoubleVector vector)

Computes the distance from the hyperplane in the kernel induced hyperspace.

Parameters:

vector - the vector to predict

Returns:

the distance from the hyperplane

predict

public double predict(DoubleVector vector)

Computes the predicted value of a vector by using the current SVM.

Parameters:

vector - the vector for which to predict the class

Returns:

the predicted value (-1 or 1)

getMargin

public double getMargin()

The margin of a SVM is the minimum distance from all support vectors to the decision hyperplane.

Returns:

margin value of SVM.

saveToPredictorParams

public void saveToPredictorParams(ModelContentWO predParams,
                                  String id)

Save the Support Vector Machine for later use.

Parameters:

predParams - where the SVM will be saved.

id - unique identifier for this SVM.

toString

public String toString()

Overrides:

toString in class Object

getPositive

public String getPositive()

Returns:

the "positive" class value.

getSupportVectors

public DoubleVector[] getSupportVectors()

Returns:

the supportVectors

getAlphas

public double[] getAlphas()

Returns:

the alpha coefficients of the SVM.

getTargetAlphas

public double[] getTargetAlphas()

Multiplies the alpha coefficients of the SVM with the target value (-1 or 1, depending if the input is a 'positive' example or not). Use this method for retrieving the alphas for a PMML compliant representation instead of the getAlphas() method.

Returns:

the alpha coefficients oft the SVM multiplied by the target factor

getThreshold

public double getThreshold()

Returns:

the threshold value b of the SVM.s