boone.links

Class MultiLink

java.lang.Object

boone.BrainPart

boone.Link

boone.links.MultiLink

All Implemented Interfaces:

Storable, java.lang.Cloneable

public class MultiLink
extends Link
implements Storable

A link used for multiple neurons. This link delegates all actions to the base link, so that all multi links with the same base link are in effect the same link. This circumvents the Boone convention that a link only has one source and one sink neuron. With the multi link a link can have multiple sources and sinks. Note that the base link must be a default Link, but may not be a custom link.

Version:

$Id$

Author:

Helmut A. Mayer

Field Summary

Fields inherited from class boone.Link

sink, source, value

Fields inherited from class boone.BrainPart

activationFn, name, NO_ID, weight

Constructor Summary

Constructors

Constructor and Description
MultiLink() Creates an empty multi link.
MultiLink(Link baseLink) Creates a link based on another link.
MultiLink(Link baseLink, Neuron source, Neuron sink, double weight) Creates a new link from source to sink with the given weight.

Method Summary

Modifier and Type	Method and Description
void	addGradient(double value) Adds a value to the gradient.
void	addToWeight(double deltaWeight) Adds the given value to the weight of the base link.
MultiLink	clone() Returns a semi-deep copy of this object.
void	fromXML(org.jdom2.Element element, NeuralNet net) Reads attributes from XML tree.
double	getAccu() Returns the accu of the base link.
Link	getBaseLink() Returns the base link of this multi link.
double	getGradient() Returns the gradient of the base link.
double	getLastGradient() Returns the last gradient of the base link.
java.lang.String	getName() Returns the name of the base link.
double	getWeight() Returns the weight of the base link.
void	propagate() Propagates the output of the given neuron over the link The weight of this link is the weight of the base link.
void	randomize() Sets the base link weight to a random value.
void	randomize(double min, double max) Sets the base link weight to a random value in [min, max).
void	reset() Resets this link and the base link.
void	setAccu(double value) Sets the accu of the base link.
void	setGradient(double gradient) Sets the gradient of the base link.
void	setLastGradient(double lastGradient) Remembers the last gradient of the base link.
void	setWeight(double weight) Sets the weight of the base link.
org.jdom2.Element	toXML(org.jdom2.Element element) Writes attributes to XML tree.

Methods inherited from class boone.Link

clearSink, clearSource, disconnect, getOtherNeuron, getSink, getSource, getValue, isDangling, isDirected, isInputOf, isOutputOf, replaceFrom, replaceNeuron, setDirected, setSink, setSource, setValue, toString

Methods inherited from class boone.BrainPart

calculate, fromXML, getActivationFn, getID, getUniqueID, isTrainable, setActivationFn, setID, setName, setTrainable

Methods inherited from class java.lang.Object

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

Methods inherited from interface boone.io.Storable

fromXML

Constructor Detail

MultiLink

public MultiLink()

Creates an empty multi link.

MultiLink

public MultiLink(Link baseLink)

Creates a link based on another link.

Parameters:

baseLink - the base link

MultiLink

public MultiLink(Link baseLink,
                 Neuron source,
                 Neuron sink,
                 double weight)

Creates a new link from source to sink with the given weight.

Parameters:

baseLink - the base link

source - the source neuron

sink - the sink neuron

weight - the weight of this link

Method Detail

clone

public MultiLink clone()

Returns a semi-deep copy of this object. Sets the source and sink neuron of the clone to null. Note that the name is taken without change from the cloned neuron. The base link is not cloned, as this is the basic idea of the multi link.

Overrides:

clone in class Link

Returns:

a semi-deep clone

getName

public java.lang.String getName()

Returns the name of the base link.

Overrides:

getName in class BrainPart

getGradient

public double getGradient()

Returns the gradient of the base link.

Overrides:

getGradient in class BrainPart

Returns:

some gradient value

setGradient

public void setGradient(double gradient)

Sets the gradient of the base link.

Overrides:

setGradient in class BrainPart

Parameters:

gradient - some gradient value

addGradient

public void addGradient(double value)

Adds a value to the gradient.

Overrides:

addGradient in class BrainPart

Parameters:

value - a value

getLastGradient

public double getLastGradient()

Returns the last gradient of the base link.

Overrides:

getLastGradient in class BrainPart

Returns:

the last gradient value

setLastGradient

public void setLastGradient(double lastGradient)

Remembers the last gradient of the base link.

Overrides:

setLastGradient in class BrainPart

Parameters:

lastGradient - the last gradient value

getAccu

public double getAccu()

Returns the accu of the base link.

Overrides:

getAccu in class BrainPart

Returns:

some accu value

setAccu

public void setAccu(double value)

Sets the accu of the base link.

Overrides:

setAccu in class BrainPart

Parameters:

value - some value

getBaseLink

public Link getBaseLink()

Returns the base link of this multi link.

Returns:

the base link

reset

public void reset()

Resets this link and the base link.

Overrides:

reset in class BrainPart

randomize

public void randomize()

Sets the base link weight to a random value.

Overrides:

randomize in class BrainPart

randomize

public void randomize(double min,
                      double max)

Sets the base link weight to a random value in [min, max).

Overrides:

randomize in class BrainPart

Parameters:

min - the minimal random weight

max - the maximal random weight

propagate

public void propagate()

Propagates the output of the given neuron over the link The weight of this link is the weight of the base link.

Overrides:

propagate in class Link

getWeight

public double getWeight()

Returns the weight of the base link.

Overrides:

getWeight in class BrainPart

Returns:

the weight value

setWeight

public void setWeight(double weight)

Sets the weight of the base link.

Overrides:

setWeight in class BrainPart

Parameters:

weight - the weight value

addToWeight

public void addToWeight(double deltaWeight)

Adds the given value to the weight of the base link.

Overrides:

addToWeight in class BrainPart

Parameters:

deltaWeight - the value to be added

toXML

public org.jdom2.Element toXML(org.jdom2.Element element)

Writes attributes to XML tree. The weight is the weight of the base link.

Specified by:

toXML in interface Storable

Overrides:

toXML in class Link

Parameters:

element - the parent element

Returns:

the link element

fromXML

public void fromXML(org.jdom2.Element element,
                    NeuralNet net)

Reads attributes from XML tree.

Overrides:

fromXML in class Link

Parameters:

element - an XML element

net - the associated net