org.cognity.neural.net

Class AbstractNeuralNet

java.lang.Object

org.cognity.neural.net.AbstractNeuralNet

All Implemented Interfaces:

Serializable, NeuralNet

Direct Known Subclasses:

AbstractFeedforwardNet

public abstract class AbstractNeuralNet
extends Object
implements NeuralNet, Serializable

Base class for all neural networks. It implements all methods that concern layer and link management, so that subclasses must implement only computional methods and the builder object.

Implementation details:
Input and output links of every layer are stored in hash maps, so that getInputLinks and getOutputLinks methods run in constant time.

Author:

Artur Markiewicz

See Also:

Serialized Form

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	AbstractNeuralNet(Layer inputLayer, Layer outputLayer, Set<Layer> layers, Set<Link> links) Creates an instance of a neural network.

Method Summary

Methods

Modifier and Type	Method and Description
Layer	getInputLayer() Returns the input layer of this network.
Set<Link>	getInputLinks(Layer layer) Returns input links of the specified layer.
int	getInputSize() Returns the size of the input layer.
int	getLayerCount() Returns the number of all layers in this network.
Set<Layer>	getLayers() Returns a set of all layers in this network.
Set<Link>	getLinks() Returns a set of all links in this network.
Link	getOneInputLink(Layer layer) Returns one of input links of the specified layer.
Link	getOneOutputLink(Layer layer) Returns one of output links of the specified layer.
Layer	getOutputLayer() Returns the output layer of this network.
Set<Link>	getOutputLinks(Layer layer) Returns output links of the specified layer.
int	getOutputSize() Returns the size of the output layer.
boolean	hasInputLinks(Layer layer) Returns true if the specified layer has at least one input link.
boolean	hasOutputLinks(Layer layer) Returns true if the specified layer has at least one output link.
protected void	initialize()
String	toString()

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.cognity.neural.net.NeuralNet

compute

Constructor Detail

AbstractNeuralNet

protected AbstractNeuralNet(Layer inputLayer,
                 Layer outputLayer,
                 Set<Layer> layers,
                 Set<Link> links)

Creates an instance of a neural network.

Parameters:

layers - the list of layers.

links - the set of links between layers.

inputLayer - the network's input layer.

outputLayer - the network's output layer.

Method Detail

getInputSize

public int getInputSize()

Description copied from interface: NeuralNet

Returns the size of the input layer.

Specified by:

getInputSize in interface NeuralNet

Returns:

the size of the input layer

getOutputSize

public int getOutputSize()

Description copied from interface: NeuralNet

Returns the size of the output layer.

Specified by:

getOutputSize in interface NeuralNet

Returns:

the size of the output layer

getInputLayer

public Layer getInputLayer()

Description copied from interface: NeuralNet

Returns the input layer of this network.

Specified by:

getInputLayer in interface NeuralNet

Returns:

the input layer of this network

getOutputLayer

public Layer getOutputLayer()

Description copied from interface: NeuralNet

Returns the output layer of this network.

Specified by:

getOutputLayer in interface NeuralNet

Returns:

the output layer of this network

hasInputLinks

public boolean hasInputLinks(Layer layer)

Description copied from interface: NeuralNet

Returns true if the specified layer has at least one input link.

Specified by:

hasInputLinks in interface NeuralNet

Parameters:

layer - the examinated layer. Cannot be null.

Returns:

• true - if specified layer has at least one input link
• false - if specified layer has no input links, doesn't belong to this network

hasOutputLinks

public boolean hasOutputLinks(Layer layer)

Description copied from interface: NeuralNet

Returns true if the specified layer has at least one output link.

Specified by:

hasOutputLinks in interface NeuralNet

Parameters:

layer - the examinated layer. Cannot be null.

Returns:

• true - if specified layer has at least one output link
• false - if specified layer has no input links or doesn't belong to this network

getInputLinks

public Set<Link> getInputLinks(Layer layer)

Description copied from interface: NeuralNet

Returns input links of the specified layer. If the layer has no input links, then empty set is returned.

The returned set is immutable.

Specified by:

getInputLinks in interface NeuralNet

Parameters:

layer - the examinated layer. Cannot be null.

Returns:

• the immutable set of input links of the layer
• empty set - if the layer doesn't belong to this network

getOutputLinks

public Set<Link> getOutputLinks(Layer layer)

Description copied from interface: NeuralNet

Returns output links of the specified layer. If the layer has no output links, then empty set is returned.

The returned set is immutable.

Specified by:

getOutputLinks in interface NeuralNet

Parameters:

layer - the layer. Cannot be null.

Returns:

• the immutable set of output links of the layer
• empty set - if the layer doesn't belong to this network

getOneInputLink

public Link getOneInputLink(Layer layer)

Description copied from interface: NeuralNet

Returns one of input links of the specified layer. If the layer has no input links, then null is returned.
This is a shortcut method. It's useful only if the layer has exactly one input link. It's not specified which link will be returned if the layer has more than one input link.

Specified by:

getOneInputLink in interface NeuralNet

Parameters:

layer - the examinated layer. Cannot be null.

Returns:

• one of input links of the layer
• null - if the layer has no input links or doesn't belong to this network

getOneOutputLink

public Link getOneOutputLink(Layer layer)

Description copied from interface: NeuralNet

Returns one of output links of the specified layer. If the layer has no output links, then null is returned.
This is a shortcut method. It's useful only if the layer has exactly one output link. It's not specified which link will be returned if the layer has more than one output link.

Specified by:

getOneOutputLink in interface NeuralNet

Parameters:

layer - the layer. Cannot be null.

Returns:

• one of output links of the layer
• null - if the layer has no output links or doesn't belong to this network

getLayers

public Set<Layer> getLayers()

Description copied from interface: NeuralNet

Returns a set of all layers in this network.

The returned list is immutable.

Specified by:

getLayers in interface NeuralNet

Returns:

the immutable set of all layers

getLinks

public Set<Link> getLinks()

Description copied from interface: NeuralNet

Returns a set of all links in this network.

The returned set is immutable.

Specified by:

getLinks in interface NeuralNet

Returns:

the immutable set of all links

getLayerCount

public int getLayerCount()

Description copied from interface: NeuralNet

Returns the number of all layers in this network.

Specified by:

getLayerCount in interface NeuralNet

Returns:

the number of layers

toString

public String toString()

Overrides:

toString in class Object

initialize

protected void initialize()