org.neuroph.util.random
Class DistortRandomizer
java.lang.Object
org.neuroph.util.random.WeightsRandomizer
org.neuroph.util.random.DistortRandomizer
public class DistortRandomizer
- extends WeightsRandomizer
This class provides distort randomization technique, which distorts existing
weight values using specified distortion factor.
Weights are distorted using following formula:
newWeightValue = currentWeightValue + (distortionFactor - (random * distortionFactor * 2))
- Author:
- Zoran Sevarac
Constructor Summary |
DistortRandomizer(double distortionFactor)
Create a new instance of DistortRandomizer with specified distortion factor |
Method Summary |
void |
randomize(NeuralNetwork neuralNetwork)
Iterate all layers, neurons and connection weight and apply distort randomization |
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
DistortRandomizer
public DistortRandomizer(double distortionFactor)
- Create a new instance of DistortRandomizer with specified distortion factor
- Parameters:
distortionFactor
- amount to distort existing weights
randomize
public void randomize(NeuralNetwork neuralNetwork)
- Iterate all layers, neurons and connection weight and apply distort randomization
- Overrides:
randomize
in class WeightsRandomizer
- Parameters:
neuralNetwork
-
Copyright © 2012. All Rights Reserved.