NearNeural

Class ReceptiveNeuronLayer

java.lang.Object

NearNeural.NeuronLayer

NearNeural.ReceptiveNeuronLayer

public class ReceptiveNeuronLayer
extends NeuronLayer

A Receptive form of the Neuron Layer, this class allows Neurons to maintain Edges to other Neurons. It has a backpropgation algorithm in it that is suitable for intermediate layers but not output. The only difference between the OutputNeuronLayer and this one is the back propagation routine, and they are separated for that reason only.

This Layer accepts any descendents of ReceptiveNeuron, but for stability reasons the Sigmoid form is coded in at the moment. Since it's the only possible algorithm that isn't exactly an issue, but be aware if you wish to change the algorithm and source later.

Version:

0.2.0a

Author:

Stephen Wattam

Field Summary

Fields inherited from class NearNeural.NeuronLayer

neurons, thresholdModel

Constructor Summary

Constructors

Constructor and Description
ReceptiveNeuronLayer(ThresholdingAlgorithm p_thresholdModel) Creates a new receptive neuron layer with the desired thresholding algorithm.

Method Summary

Methods

Modifier and Type	Method and Description
Edge	getEdge(int p_neuronIndex, int p_edgeIndex) Returns an edge from one of the neurons in this layer.
ReceptiveNeuron	getNeuron(int p_index) Returns a neuron with the index provided
ReceptiveNeuron[]	getNeurons() Returns all neurons as ReceptiveNeurons.
protected void	linkTo(NeuronLayer p_neuronlayer) Links all neurons in this layer to all neurons in the layer provided.
void	preCalculate() Precalculates all neuron values based on their edges.

Methods inherited from class NearNeural.NeuronLayer

addNeuron, countNeurons, isPresent, removeNeuron, removeNeuron

Methods inherited from class java.lang.Object

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

Constructor Detail

ReceptiveNeuronLayer

public ReceptiveNeuronLayer(ThresholdingAlgorithm p_thresholdModel)

Creates a new receptive neuron layer with the desired thresholding algorithm.

Parameters:

p_thresholdModel - The thresholding algorithm to use./

Method Detail

linkTo

protected void linkTo(NeuronLayer p_neuronlayer)

Links all neurons in this layer to all neurons in the layer provided.

Parameters:

p_neuronlayer - The neuron layer to link all Neurons to.

getEdge

public Edge getEdge(int p_neuronIndex,
           int p_edgeIndex)
             throws IndexOutOfBoundsException

Returns an edge from one of the neurons in this layer.

Parameters:

p_neuronIndex - The index of the neron from which to return an edge.

p_edgeIndex - The index of the edge from that neuron

Returns:

The edge as selected by the above two indices

Throws:

IndexOutOfBoundsException - in the event that either of the indices fall out of range

preCalculate

public void preCalculate()

Precalculates all neuron values based on their edges. It is imperative that this is run in order, however, if you only change one layer it is quite possible to run only this method.

getNeuron

public ReceptiveNeuron getNeuron(int p_index)
                          throws IndexOutOfBoundsException

Description copied from class: NeuronLayer

Returns a neuron with the index provided

Overrides:

getNeuron in class NeuronLayer

Parameters:

p_index - The index of the Neuron to return

Returns:

The neuron at the index specified

Throws:

IndexOutOfBoundsException - in the event that the index is out of range.

getNeurons

public ReceptiveNeuron[] getNeurons()

Returns all neurons as ReceptiveNeurons. Since this is a receptive neuron layer we can cast down with relative safety. This is used for back propagation.

Overrides:

getNeurons in class NeuronLayer

Returns:

All of the contained neurons, cast to ReceptiveNeurons