gov.sandia.cognition.learning.algorithm.svm

Class SuccessiveOverrelaxation.Entry

java.lang.Object

gov.sandia.cognition.util.AbstractCloneableSerializable

gov.sandia.cognition.util.AbstractWeighted

gov.sandia.cognition.util.DefaultWeightedValue<InputType>

gov.sandia.cognition.learning.algorithm.svm.SuccessiveOverrelaxation.Entry

All Implemented Interfaces:

CloneableSerializable, Weighted, WeightedValue<InputType>, java.io.Serializable, java.lang.Cloneable, java.lang.Comparable<SuccessiveOverrelaxation.Entry>

Enclosing class:

SuccessiveOverrelaxation<InputType>

protected class SuccessiveOverrelaxation.Entry
extends DefaultWeightedValue<InputType>
implements java.lang.Comparable<SuccessiveOverrelaxation.Entry>

The Entry class represents the data that the algorithm keeps about each training example.

See Also:

Serialized Form

Nested Class Summary

Nested classes/interfaces inherited from class gov.sandia.cognition.util.DefaultWeightedValue

DefaultWeightedValue.WeightComparator

Field Summary

Fields

Modifier and Type	Field and Description
protected InputOutputPair<? extends InputType,? extends java.lang.Boolean>	example The example the data pertains to.
protected boolean	output The output represented as a raw boolean, to enforce that the label exists.
protected double	outputDouble The output converted to a double form (+1.0 or -1.0).
protected double	previousStepWeight The weight of the entry on the previous step.
protected double	selfKernel This is the value of the kernel applied to the example and itself.
protected boolean	supportInserted Indicates if the support vector has been inserted into the map of support vectors or not.

Fields inherited from class gov.sandia.cognition.util.DefaultWeightedValue

value

Fields inherited from class gov.sandia.cognition.util.AbstractWeighted

DEFAULT_WEIGHT, weight

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	Entry(InputOutputPair<? extends InputType,? extends java.lang.Boolean> example) Creates a new Entry for the given example.

Method Summary

Modifier and Type	Method and Description
int	compareTo(SuccessiveOverrelaxation.Entry other) Compares this entry to another one by comparing the weights.
InputType	getInput() Gets the input that the data belongs to.
boolean	getOutput() Gets the output value of the entry as a boolean.
double	getUnlabeledWeight() Gets the unlabeled weight.
void	setUnlabeledWeight(double unlabeledWeight) Sets the unlabeled weight.

Methods inherited from class gov.sandia.cognition.util.DefaultWeightedValue

clone, create, getValue, setValue

Methods inherited from class gov.sandia.cognition.util.AbstractWeighted

getWeight, setWeight

Methods inherited from class java.lang.Object

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

Methods inherited from interface gov.sandia.cognition.util.Weighted

getWeight

Field Detail

example

protected InputOutputPair<? extends InputType,? extends java.lang.Boolean> example

The example the data pertains to.

output

protected boolean output

The output represented as a raw boolean, to enforce that the label exists.

outputDouble

protected double outputDouble

The output converted to a double form (+1.0 or -1.0).

supportInserted

protected boolean supportInserted

Indicates if the support vector has been inserted into the map of support vectors or not. This allows us to keep the supports map to only contain the entries whose weights are non-zero.

selfKernel

protected double selfKernel

This is the value of the kernel applied to the example and itself. We use this value in the update step, so we can cache it for a performance boost.

previousStepWeight

protected double previousStepWeight

The weight of the entry on the previous step. This is used at the end of the step to calculate the total change of weights in the step.

Constructor Detail

Entry

protected Entry(InputOutputPair<? extends InputType,? extends java.lang.Boolean> example)

Creates a new Entry for the given example.

Parameters:

example - The example to create the entry for.

Method Detail

getInput

public InputType getInput()

Gets the input that the data belongs to.

Returns:

The input.

getOutput

public boolean getOutput()

Gets the output value of the entry as a boolean.

Returns:

The output.

setUnlabeledWeight

public void setUnlabeledWeight(double unlabeledWeight)

Sets the unlabeled weight. This means that the label is applied to the weight when it is set. Must be non-negative.

Parameters:

unlabeledWeight - The unlabeled weight.

getUnlabeledWeight

public double getUnlabeledWeight()

Gets the unlabeled weight. This means that the label part of the weight value is removed. This means a non-negative value is returned.

Returns:

The unlabeled weight.

compareTo

public int compareTo(SuccessiveOverrelaxation.Entry other)

Compares this entry to another one by comparing the weights.

Specified by:

compareTo in interface java.lang.Comparable<SuccessiveOverrelaxation.Entry>

Parameters:

other - The entry to compare to.

Returns:

The comparison based on weight.