org.simbrain.network.subnetworks

Class CompetitiveGroup

java.lang.Object

org.simbrain.network.groups.Group

org.simbrain.network.groups.NeuronGroup

org.simbrain.network.subnetworks.CompetitiveGroup

All Implemented Interfaces:

CopyableGroup<NeuronGroup>

public class CompetitiveGroup
extends NeuronGroup

Competitive implements a simple competitive network. Current implementations include Rummelhart-Zipser (PDP, 151-193), and Alvarez-Squire 1994, PNAS, 7041-7045.

Author:

Jeff Yoshimi

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	CompetitiveGroup.UpdateMethod Specific implementation of competitive learning.

Field Summary

Fields inherited from class org.simbrain.network.groups.NeuronGroup

DEFAULT_GROUP_SIZE, DEFAULT_LAYOUT, FLUSH_FREQUENCY

Constructor Summary

Constructors

Constructor and Description
CompetitiveGroup(Network newRoot, CompetitiveGroup oldNet) Copy constructor.
CompetitiveGroup(Network root, int numNeurons) Constructs a competitive network with specified number of neurons.

Method Summary

Modifier and Type	Method and Description
CompetitiveGroup	deepCopy(Network newParent) Returns a deep copy of a copyable group using the same root network as the original.
double	getLeakyLearningRate() Return leaky learning rate.
double	getLearningRate() Return the learning rate.
double	getLoseValue() Return the loser value.
boolean	getNormalizeInputs() Return the normalize inputs value.
double	getSynpaseDecayPercent()
java.lang.String	getTypeDescription() Return a human-readable name for this type of neuron group.
CompetitiveGroup.UpdateMethod	getUpdateMethod()
boolean	getUseLeakyLearning() Return the leaky learning value.
double	getWinValue() Return the winner value.
void	normalizeAllIncomingWeights() Normalize all weights coming in to this network.
void	normalizeIncomingWeights() Normalize weights coming in to this network, separately for each neuron.
void	randomize() Randomize and normalize weights.
void	randomizeIncomingWeights() Randomize all weights coming in to this network.
void	setLeakyLearningRate(double leakyRate) Sets the leaky learning rate.
void	setLearningRate(double rate) Sets learning rate.
void	setLoseValue(double loseValue) Sets the loser value.
void	setNormalizeInputs(boolean normalizeInputs) Sets the normalize inputs value.
void	setSynpaseDecayPercent(double synpaseDecayPercent)
void	setUpdateMethod(CompetitiveGroup.UpdateMethod updateMethod)
void	setUseLeakyLearning(boolean useLeakyLearning) Sets the leaky learning value.
void	setWinValue(double winValue) Sets the winner value.
void	update() Updates all the neurons in the neuron group according to their NeuronUpdateRule(s).

Methods inherited from class org.simbrain.network.groups.NeuronGroup

addIncomingSg, addNeuron, addNeuron, addOutgoingSg, applyInputs, applyLayout, applyLayout, clearActivations, clearNeuronList, containsAsIncoming, containsAsOutgoing, containsNeuron, copyActivations, delete, forceSetActivationLevels, forceSetActivations, getActivations, getBetweenNeuronInterval, getBiases, getCenterX, getCenterY, getExternalActivations, getFourCorners, getGridThreshold, getHeight, getIncomingSgs, getIncomingWeights, getLabelsOfActiveNeurons, getLayout, getMaxDim, getMaxX, getMaxY, getMinX, getMinY, getMostActiveNeuron, getNeuronList, getNeuronsInRadius, getNeuronType, getNumSubSamples, getOutgoingSg, getOutgoingWeights, getPosition, getSpikeIndexes, getTestData, getUpdateMethodDesecription, getWidth, inFanInOfSomeNode, isAllClamped, isAllUnclamped, isEmpty, isInputMode, isRecordAsSpikes, isRecording, isSpikingNeuronGroup, isUseSubSampling, offset, printActivations, prune, randomizeBiases, randomizeOutgoingWeights, readNextInputs, removeIncomingSg, removeNeuron, removeOutgoingSg, resetSubsamplingIndices, setActivationLevels, setActivations, setBetweenNeuronInterval, setClamped, setGridThreshold, setIncrement, setInputMode, setInputValues, setLayout, setLayoutBasedOnSize, setLayoutBasedOnSize, setLocation, setLowerBound, setNeuronType, setNeuronType, setNumSubSamples, setRecordAsSpikes, setSpikingNeuronGroup, setTestData, setUpperBound, setUseSubSampling, setXYZCoordinatesFromFile, size, startRecording, stopRecording, testAndSetIfSpiking, toString, writeActsToFile

Methods inherited from class org.simbrain.network.groups.Group

getId, getLabel, getParentGroup, getParentNetwork, getStateInfo, hasParentGroup, isMarkedForDeletion, isTopLevelGroup, recursivelySetIds, setId, setLabel, setMarkedForDeletion, setParentGroup, setStateInfo

Methods inherited from class java.lang.Object

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

Constructor Detail

CompetitiveGroup

public CompetitiveGroup(Network root,
                        int numNeurons)

Constructs a competitive network with specified number of neurons.

Parameters:

numNeurons - size of this network in neurons

root - reference to Network.

CompetitiveGroup

public CompetitiveGroup(Network newRoot,
                        CompetitiveGroup oldNet)

Copy constructor.

Parameters:

newRoot - new root network

oldNet - old network.

Method Detail

deepCopy

public CompetitiveGroup deepCopy(Network newParent)

Description copied from interface: CopyableGroup

Returns a deep copy of a copyable group using the same root network as the original.

Specified by:

deepCopy in interface CopyableGroup<NeuronGroup>

Overrides:

deepCopy in class NeuronGroup

Parameters:

newParent - the parent network for this group, potentially different from the original (used when copying and pasting from one network to another)

Returns:

a deep copy of a group

getTypeDescription

public java.lang.String getTypeDescription()

Description copied from class: NeuronGroup

Return a human-readable name for this type of neuron group. Subclasses should override this. Used in the Gui for various purposes.

Overrides:

getTypeDescription in class NeuronGroup

Returns:

the name of this type of neuron group.

update

public void update()

Description copied from class: NeuronGroup

Updates all the neurons in the neuron group according to their NeuronUpdateRule(s). If the group is in input mode reads in the next set of values from the input table and sets the neuron values accordingly.

Overrides:

update in class NeuronGroup

normalizeIncomingWeights

public void normalizeIncomingWeights()

Normalize weights coming in to this network, separately for each neuron.

normalizeAllIncomingWeights

public void normalizeAllIncomingWeights()

Normalize all weights coming in to this network.

randomizeIncomingWeights

public void randomizeIncomingWeights()

Randomize all weights coming in to this network. TODO: Add gaussian option...

Overrides:

randomizeIncomingWeights in class NeuronGroup

randomize

public void randomize()

Randomize and normalize weights.

Overrides:

randomize in class NeuronGroup

getLearningRate

public double getLearningRate()

Return the learning rate.

Returns:

the learning rate

setLearningRate

public void setLearningRate(double rate)

Sets learning rate.

Parameters:

rate - The new epsilon value.

getLoseValue

public final double getLoseValue()

Return the loser value.

Returns:

the loser Value

setLoseValue

public final void setLoseValue(double loseValue)

Sets the loser value.

Parameters:

loseValue - The new loser value

getWinValue

public final double getWinValue()

Return the winner value.

Returns:

the winner value

setWinValue

public final void setWinValue(double winValue)

Sets the winner value.

Parameters:

winValue - The new winner value

getLeakyLearningRate

public double getLeakyLearningRate()

Return leaky learning rate.

Returns:

Leaky learning rate

setLeakyLearningRate

public void setLeakyLearningRate(double leakyRate)

Sets the leaky learning rate.

Parameters:

leakyRate - Leaky rate value to set

getNormalizeInputs

public boolean getNormalizeInputs()

Return the normalize inputs value.

Returns:

the normailize inputs value

setNormalizeInputs

public void setNormalizeInputs(boolean normalizeInputs)

Sets the normalize inputs value.

Parameters:

normalizeInputs - Normalize inputs value to set

getUseLeakyLearning

public boolean getUseLeakyLearning()

Return the leaky learning value.

Returns:

the leaky learning value

setUseLeakyLearning

public void setUseLeakyLearning(boolean useLeakyLearning)

Sets the leaky learning value.

Parameters:

useLeakyLearning - The leaky learning value to set

getSynpaseDecayPercent

public double getSynpaseDecayPercent()

Returns:

the synpaseDecayPercent

setSynpaseDecayPercent

public void setSynpaseDecayPercent(double synpaseDecayPercent)

Parameters:

synpaseDecayPercent - the synpaseDecayPercent to set

getUpdateMethod

public CompetitiveGroup.UpdateMethod getUpdateMethod()

Returns:

the updateMethod

setUpdateMethod

public void setUpdateMethod(CompetitiveGroup.UpdateMethod updateMethod)

Parameters:

updateMethod - the updateMethod to set