org.ea.javacnn.trainers

Class WindowGradTrainer

java.lang.Object

org.ea.javacnn.trainers.Trainer

org.ea.javacnn.trainers.WindowGradTrainer

public class WindowGradTrainer
extends Trainer

This is AdaGrad but with a moving window weighted average so the gradient is not accumulated over the entire history of the run. it's also referred to as Idea #1 in Zeiler paper on AdaDelta.

Constructor Summary

Constructors

Constructor and Description
WindowGradTrainer(JavaCNN net, int batch_size, float l2_decay)

Method Summary

Modifier and Type	Method and Description
void	update(int i, int j, double gij, double[] p)

Methods inherited from class org.ea.javacnn.trainers.Trainer

initTrainData, train

Methods inherited from class java.lang.Object

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

Constructor Detail

WindowGradTrainer

public WindowGradTrainer(JavaCNN net,
                         int batch_size,
                         float l2_decay)

Method Detail

update

public void update(int i,
                   int j,
                   double gij,
                   double[] p)

Specified by:

update in class Trainer