mulan.classifier.neural.model
Class BasicNeuralNet

java.lang.Object
  mulan.classifier.neural.model.BasicNeuralNet

All Implemented Interfaces:

Serializable, NeuralNet

public class BasicNeuralNet
extends Object
implements NeuralNet, Serializable

Implementation of basic neural network. The network consists of one input layer, zero or more hidden layers and one output layer. Each layer contains 1 or more Neuron units. The input layer is used just to store and forward input pattern of the network to first hidden layer for processing. Input layer do not process input pattern. Neurons of input layer have one input weight equal to 1, bias weight equal to 0 and use linear activation function.

Version:

2012.02.27

Author:

Jozef Vilcek

See Also:

Serialized Form

Constructor Summary

BasicNeuralNet(int[] netTopology, double biasInput, Class<? extends ActivationFunction> activationFunction, Random random)
Creates a new BasicNeuralNet instance.

Method Summary

double[]	feedForward(double[] inputPattern) Propagates the input pattern through the network.
int	getLayersCount() Returns a total number of layers of the neural network.
List<Neuron>	getLayerUnits(int layerIndex) Returns units of a particular layer of the neural network.
int	getNetInputSize() Gets the size/dimension of the input layer of the neural network.
int	getNetOutputSize() Gets the size/dimension of the output layer of the neural network.
double[]	getOutput() Returns the actual output of the neural network, which is a result of last processed input pattern.
void	reset() Perform reset, re-initialization of neural network.

Methods inherited from class java.lang.Object

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

Constructor Detail

BasicNeuralNet

public BasicNeuralNet(int[] netTopology,
                      double biasInput,
                      Class<? extends ActivationFunction> activationFunction,
                      Random random)

Creates a new BasicNeuralNet instance.

Parameters:

netTopology - defines a topology of the network. The array length corresponds to number of network layers. The values of the array corresponds to number of neurons in each particular layer.

biasInput - the bias input value for neurons of the neural network.

activationFunction - the type of activation function to be used by network elements

random - the pseudo-random generator instance to be used for computations involving randomness. This parameter can be null. In this case, new random instance with default seed will be constructed where needed.

Throws:

IllegalArgumentException - if network topology is incorrect of activation function class is null.

Method Detail

getLayerUnits

public List<Neuron> getLayerUnits(int layerIndex)

Description copied from interface: NeuralNet

Returns units of a particular layer of the neural network. The valid indexes for layers are from 0 to N-1, where N is total number of layers
The first layer (index = 0) is always input layer and last (index = N-1) always output layer.

Specified by:

getLayerUnits in interface NeuralNet

Returns:

returns an unmodifiable list of units of the particular layer

See Also:

Collections.unmodifiableList(List)

getLayersCount

public int getLayersCount()

Description copied from interface: NeuralNet

Returns a total number of layers of the neural network.

Specified by:

getLayersCount in interface NeuralNet

Returns:

the number of layers in the neural network

feedForward

public double[] feedForward(double[] inputPattern)

Description copied from interface: NeuralNet

Propagates the input pattern through the network.

Specified by:

feedForward in interface NeuralNet

Parameters:

inputPattern - the input pattern for the network to process

Returns:

the output of the network

getOutput

public double[] getOutput()

Description copied from interface: NeuralNet

Returns the actual output of the neural network, which is a result of last processed input pattern.

Specified by:

getOutput in interface NeuralNet

Returns:

the output of the network. Returns null if network is reset or no input pattern was processed

reset

public void reset()

Description copied from interface: NeuralNet

Perform reset, re-initialization of neural network. All learned knowledge stored in the network will be lost.

Specified by:

reset in interface NeuralNet

getNetInputSize

public int getNetInputSize()

Description copied from interface: NeuralNet

Gets the size/dimension of the input layer of the neural network. This is the size of input pattern the neural network can process.

Specified by:

getNetInputSize in interface NeuralNet

Returns:

the network input size

getNetOutputSize

public int getNetOutputSize()

Description copied from interface: NeuralNet

Gets the size/dimension of the output layer of the neural network. This is the size of output pattern the neural network produces.

Specified by:

getNetOutputSize in interface NeuralNet

Returns:

the network output size