org.apache.mahout.classifier.sgd
Class MixedGradient

java.lang.Object
  extended by org.apache.mahout.classifier.sgd.MixedGradient
All Implemented Interfaces:
Gradient

public class MixedGradient
extends Object
implements Gradient

Provides a stochastic mixture of ranking updates and normal logistic updates. This uses a combination of AUC driven learning to improve ranking performance and traditional log-loss driven learning to improve log-likelihood.

See www.eecs.tufts.edu/~dsculley/papers/combined-ranking-and-regression.pdf

This implementation only makes sense for the binomial case.


Constructor Summary
MixedGradient(double alpha, int window)
           
 
Method Summary
 Vector apply(String groupKey, int actual, Vector instance, AbstractVectorClassifier classifier)
           
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Constructor Detail

MixedGradient

public MixedGradient(double alpha,
                     int window)
Method Detail

apply

public Vector apply(String groupKey,
                    int actual,
                    Vector instance,
                    AbstractVectorClassifier classifier)
Specified by:
apply in interface Gradient


Copyright © 2008–2014 The Apache Software Foundation. All rights reserved.