org.encog.mathutil.randomize
Class NguyenWidrowRandomizer

java.lang.Object
  org.encog.mathutil.randomize.BasicRandomizer
      org.encog.mathutil.randomize.RangeRandomizer
          org.encog.mathutil.randomize.NguyenWidrowRandomizer

All Implemented Interfaces:

Randomizer

public class NguyenWidrowRandomizer
extends RangeRandomizer
implements Randomizer

Implementation of Nguyen-Widrow weight initialization. This is the default weight initialization used by Encog, as it generally provides the most trainable neural network.

Author:

St?phan Corriveau

Constructor Summary

NguyenWidrowRandomizer(double min, double max)
Construct a Nguyen-Widrow randomizer.

Method Summary

void	randomize(BasicNetwork network, int fromLayer) Randomize one level of a neural network.
void	randomize(MLMethod method) The Nguyen-Widrow initialization algorithm is the following : 1.

Methods inherited from class org.encog.mathutil.randomize.RangeRandomizer

getMax, getMin, randomInt, randomize, randomize

Methods inherited from class org.encog.mathutil.randomize.BasicRandomizer

getRandom, nextDouble, nextDouble, randomize, randomize, randomize, randomize, setRandom

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.encog.mathutil.randomize.Randomizer

randomize, randomize, randomize, randomize, randomize

Constructor Detail

NguyenWidrowRandomizer

public NguyenWidrowRandomizer(double min,
                              double max)

Construct a Nguyen-Widrow randomizer.

Parameters:

min - The min of the range.

max - The max of the range.

Method Detail

randomize

public final void randomize(MLMethod method)

The Nguyen-Widrow initialization algorithm is the following :
1. Initialize all weight of hidden layers with (ranged) random values
2. For each hidden layer
2.1 calculate beta value, 0.7 * Nth(#neurons of input layer) root of #neurons of current layer
2.2 for each synapse
2.1.1 for each weight
2.1.2 Adjust weight by dividing by norm of weight for neuron and multiplying by beta value

Specified by:

randomize in interface Randomizer

Overrides:

randomize in class BasicRandomizer

Parameters:

method - The network to randomize.

randomize

public void randomize(BasicNetwork network,
                      int fromLayer)

Randomize one level of a neural network.

Overrides:

randomize in class BasicRandomizer

Parameters:

network - The network to randomize

fromLayer - The from level to randomize.