org.jscience.computing.ai.neurocomputing
Class FFNLayer

java.lang.Object
  org.jscience.computing.ai.neurocomputing.FFNLayer

Direct Known Subclasses:

FFNHiddenLayer, FFNInputLayer, FFNOutputLayer

public abstract class FFNLayer
extends java.lang.Object

An abstract class for the feedforward layers.

Field Summary

protected Activation	activate The activation function this layer uses.
protected FFNLayer	connected The next layer that this layer is connected to.
protected FeedForwardNN	myNetwork The network this layer belongs to.
protected double[]	outputs The outputs for this layer.
protected double[][]	weights The weights for this layer.

Constructor Summary

FFNLayer()
Creates a new instance of FFNLayer

Method Summary

void	connectWith(FFNLayer connect) Connects this layer with another layer.
double[]	getOutput() Retrieve the outputs for this layer.
abstract int	getUnitCount() Get the number of units in this layer.
void	randomizeWeights(double min, double max) Randomize the weights between the given range.
void	run(double[] input) Run this layer on the given inputs.
void	setNetwork(FeedForwardNN network) Set the network for this layer.
void	setWeights(double[][] newWeights) Set the weights for this layer.
void	setWeights(int unit, double[] newWeights) Set the weights for a given units.

Methods inherited from class java.lang.Object

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

Field Detail

outputs

protected double[] outputs

The outputs for this layer.

weights

protected double[][] weights

The weights for this layer.

connected

protected FFNLayer connected

The next layer that this layer is connected to.

myNetwork

protected FeedForwardNN myNetwork

The network this layer belongs to.

activate

protected Activation activate

The activation function this layer uses.

Constructor Detail

FFNLayer

public FFNLayer()

Creates a new instance of FFNLayer

Method Detail

setNetwork

public void setNetwork(FeedForwardNN network)

Set the network for this layer.

Parameters:

network - the network to be used.

connectWith

public void connectWith(FFNLayer connect)

Connects this layer with another layer. This method is should normally only be called FeedForwardNN.initialize.

Parameters:

connect - the connecting layer.

See Also:

FeedForwardNN.initialize()

setWeights

public void setWeights(double[][] newWeights)

Set the weights for this layer. This method iteratively calls: for (int u=0; u

Parameters:

newWeights - the new weights.

setWeights

public void setWeights(int unit,
                       double[] newWeights)

Set the weights for a given units.

Parameters:

unit - the unit to set the weights for.

newWeights - the new weight array.

randomizeWeights

public void randomizeWeights(double min,
                             double max)

Randomize the weights between the given range.

Parameters:

min - the range minimum.

max - the range maximum.

getUnitCount

public abstract int getUnitCount()

Get the number of units in this layer.

Returns:

the number of units.

getOutput

public double[] getOutput()

Retrieve the outputs for this layer.

Returns:

the layer output.

run

public void run(double[] input)
         throws NeuralNetworkException

Run this layer on the given inputs.

Parameters:

input - the inputs for this layer.

Throws:

NeuralNetworkException - unused.