org.apache.mahout.classifier.sgd

Class RankingGradient

java.lang.Object

org.apache.mahout.classifier.sgd.RankingGradient

All Implemented Interfaces:

Gradient

public class RankingGradient
extends Object
implements Gradient

Uses the difference between this instance and recent history to get a gradient that optimizes ranking performance. Essentially this is the same as directly optimizing AUC. It isn't expected that this would be used alone, but rather that a MixedGradient would use it and a DefaultGradient together to combine both ranking and log-likelihood goals.

Constructor Summary

Constructors

Constructor and Description
RankingGradient(int window)

Method Summary

Modifier and Type	Method and Description
void	addToHistory(int actual, Vector instance)
Vector	apply(String groupKey, int actual, Vector instance, AbstractVectorClassifier classifier)
Gradient	getBaseGradient()

Methods inherited from class java.lang.Object

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

Constructor Detail

RankingGradient

public RankingGradient(int window)

Method Detail

apply

public final Vector apply(String groupKey,
                          int actual,
                          Vector instance,
                          AbstractVectorClassifier classifier)

Specified by:

apply in interface Gradient

addToHistory

public void addToHistory(int actual,
                         Vector instance)

getBaseGradient

public Gradient getBaseGradient()