Package libai.nn.supervised

Class Hebb

java.lang.Object

libai.nn.NeuralNetwork

libai.nn.supervised.SupervisedLearning

libai.nn.supervised.Hebb

All Implemented Interfaces:

java.io.Serializable

public class Hebb
extends SupervisedLearning

Hebbian supervised networks are good for pattern retrieval and reconstructions. These networks are only able to learn binary patterns because its output function (symmetric sign). However, they can deal with partially corrupted patterns and retrieve the original one without noise. The Hebbian networks uses the Hebb's rule for training. The Hebb's rule is one of the most important training rules in unsupervised networks. Other algorithms like Kohonen uses this rule as base.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
protected double	phi
protected static SymmetricSign	sign
protected Matrix	W

Fields inherited from class libai.nn.NeuralNetwork

plotter, progress, random

Constructor Summary

Constructors

Constructor	Description
Hebb(int inputs, int outputs)	Constructor.
Hebb(int inputs, int outputs, double phi)	Constructor.
Hebb(int inputs, int outputs, double phi, java.util.Random rand)	Constructor.

Method Summary

Modifier and Type	Method	Description
Matrix	getWeights()
Column	simulate(Column pattern)	Calculates the output for the pattern.
void	simulate(Column pattern, Column result)	Calculate the output for the pattern and left the result on result.
void	train(Column[] patterns, Column[] answers, double alpha, int epochs, int offset, int length, double minerror)	Train the network using the Hebb's rule with decay.

Methods inherited from class libai.nn.NeuralNetwork

error, error, euclideanDistance2, euclideanDistance2, gaussian, getDefaultRandomGenerator, getPlotter, getProgressBar, initializeProgressBar, open, open, open, save, setPlotter, setProgressBar, train, train

Methods inherited from class java.lang.Object

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

Methods inherited from class libai.nn.supervised.SupervisedLearning

validatePreconditions

Field Detail

sign

protected static final SymmetricSign sign

phi

protected double phi

W

protected Matrix W

Constructor Detail

Hebb

public Hebb(int inputs,
            int outputs)

Constructor. Creates a Hebbian network with the equals number of inputs and outputs. Set the decay constant to zero to eliminate it. Alias of this(inputs, 0);

Parameters:

inputs - Number of inputs for the network.

outputs - Number of outputs for the network.

Hebb

public Hebb(int inputs,
            int outputs,
            double phi)

Constructor. Creates a Hebbian network with the equals number of inputs and outputs. Set the constant for decay phi. If phi = 0 the network don't forget anything, if phi = 1 the network just remember the las pattern.

Parameters:

inputs - Number of inputs and outputs for the networks.

outputs - Number of outputs for the network.

phi - Decay constant.

Hebb

public Hebb(int inputs,
            int outputs,
            double phi,
            java.util.Random rand)

Constructor. Creates a Hebbian network with the equals number of inputs and outputs. Set the constant for decay phi. If phi = 0 the network don't forget anything, if phi = 1 the network just remember the las pattern.

Parameters:

inputs - Number of inputs and outputs for the networks.

outputs - Number of outputs for the network.

phi - Decay constant.

rand - Random generator used for creating matrices

Method Detail

train

public void train(Column[] patterns,
                  Column[] answers,
                  double alpha,
                  int epochs,
                  int offset,
                  int length,
                  double minerror)

Train the network using the Hebb's rule with decay. The hebb's rule, consist on reinforce the right connections if they produce a correct answer and inhibit the others. The decay term has an influence in how much affects the previous knowledge to the reinforcement.

Specified by:

train in class NeuralNetwork

Parameters:

patterns - The patterns to be learned.

answers - The expected answers.

alpha - The learning rate.

epochs - The maximum number of iterations

offset - The first pattern position

length - How many patterns will be used.

minerror - The minimal error expected.

simulate

public Column simulate(Column pattern)

Description copied from class: NeuralNetwork

Calculates the output for the pattern.

Specified by:

simulate in class NeuralNetwork

Parameters:

pattern - Pattern to use as input.

Returns:

The output for the neural network.

simulate

public void simulate(Column pattern,
                     Column result)

Calculate the output for the pattern and left the result on result. result = sign(W * pattern)

Specified by:

simulate in class NeuralNetwork

Parameters:

pattern - The input pattern

result - The output result.

getWeights

public Matrix getWeights()