org.encog.neural.neat

Class NEATPopulation

java.lang.Object

org.encog.ml.BasicML

org.encog.ml.ea.population.BasicPopulation

org.encog.neural.neat.NEATPopulation

All Implemented Interfaces:

java.io.Serializable, Population, MLError, MLInput, MLInputOutput, MLMethod, MLOutput, MLProperties, MLRegression

public class NEATPopulation
extends BasicPopulation
implements java.io.Serializable, MLError, MLRegression

A population for a NEAT or HyperNEAT system. This population holds the genomes, substrate and other values for a NEAT or HyperNEAT network. ----------------------------------------------------------------------------- http://www.cs.ucf.edu/~kstanley/ Encog's NEAT implementation was drawn from the following three Journal Articles. For more complete BibTeX sources, see NEATNetwork.java. Evolving Neural Networks Through Augmenting Topologies Generating Large-Scale Neural Networks Through Discovering Geometric Regularities Automatic feature selection in neuroevolution

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
static int	DEFAULT_CYCLES Default number of activation cycles.
static double	DEFAULT_HYPERNEAT_WEIGHT_RANGE Default link weight range for HyperNEAT networks.
static double	DEFAULT_NEAT_WEIGHT_RANGE Default link weight range for NEAT networks.
static double	DEFAULT_SURVIVAL_RATE The default survival rate.
static java.lang.String	PROPERTY_CYCLES Property to hold the number of cycles.
static java.lang.String	PROPERTY_NEAT_ACTIVATION The activation function to use.
static java.lang.String	PROPERTY_POPULATION_SIZE Property tag for the population size.
static java.lang.String	PROPERTY_SURVIVAL_RATE Property tag for the survival rate.

Constructor Summary

Constructors

Constructor and Description
NEATPopulation() An empty constructor for serialization.
NEATPopulation(int inputCount, int outputCount, int populationSize) Construct a starting NEAT population.
NEATPopulation(Substrate theSubstrate, int populationSize) Construct a starting HyperNEAT population.

Method Summary

Modifier and Type	Method and Description
long	assignGeneID()
long	assignInnovationID()
double	calculateError(MLDataSet data) Calculate the error of the ML method, given a dataset.
static double	clampWeight(double w, double weightRange) Change the weight, do not allow the weight to go out of the weight range.
MLData	compute(MLData input) Compute regression.
int	getActivationCycles()
ChooseObject<ActivationFunction>	getActivationFunctions()
GeneticCODEC	getCODEC()
GenerateID	getGeneIDGenerate()
NEATGenomeFactory	getGenomeFactory()
double	getInitialConnectionDensity()
GenerateID	getInnovationIDGenerate()
NEATInnovationList	getInnovations()
int	getInputCount()
int	getOutputCount()
RandomFactory	getRandomNumberFactory()
Substrate	getSubstrate()
double	getSurvivalRate()
double	getWeightRange()
boolean	isHyperNEAT()
void	reset() Create an initial random population.
void	setActivationCycles(int activationCycles) Set the number of activation cycles to use.
void	setCODEC(GeneticCODEC codec)
void	setInitialConnectionDensity(double initialConnectionDensity)
void	setInnovations(NEATInnovationList theInnovations) Set the innovation list to use.
void	setInputCount(int inputCount)
void	setNEATActivationFunction(ActivationFunction af) Specify to use a single activation function.
void	setOutputCount(int outputCount)
void	setRandomNumberFactory(RandomFactory randomNumberFactory)
void	setSubstrate(Substrate substrate)
void	setSurvivalRate(double theSurvivalRate) Set the survival rate, this is the percent of the population allowed to mate.
void	setWeightRange(double weightRange) Sets the weight range for links in genomes in this population.

Methods inherited from class org.encog.ml.ea.population.BasicPopulation

clear, createSpecies, determineBestSpecies, flatten, getBestGenome, getMaxIndividualSize, getName, getPopulationSize, getRules, getSpecies, purgeInvalidGenomes, setBestGenome, setGenomeFactory, setName, setPopulationSize, setRules, size, updateProperties

Methods inherited from class org.encog.ml.BasicML

getProperties, getPropertyDouble, getPropertyLong, getPropertyString, setProperty, setProperty, setProperty

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_SURVIVAL_RATE

public static final double DEFAULT_SURVIVAL_RATE

The default survival rate.

See Also:

Constant Field Values

PROPERTY_NEAT_ACTIVATION

public static final java.lang.String PROPERTY_NEAT_ACTIVATION

The activation function to use.

See Also:

Constant Field Values

PROPERTY_POPULATION_SIZE

public static final java.lang.String PROPERTY_POPULATION_SIZE

Property tag for the population size.

See Also:

Constant Field Values

PROPERTY_SURVIVAL_RATE

public static final java.lang.String PROPERTY_SURVIVAL_RATE

Property tag for the survival rate.

See Also:

Constant Field Values

DEFAULT_CYCLES

public static final int DEFAULT_CYCLES

Default number of activation cycles.

See Also:

Constant Field Values

PROPERTY_CYCLES

public static final java.lang.String PROPERTY_CYCLES

Property to hold the number of cycles.

See Also:

Constant Field Values

DEFAULT_NEAT_WEIGHT_RANGE

public static final double DEFAULT_NEAT_WEIGHT_RANGE

Default link weight range for NEAT networks.

See Also:

Constant Field Values

DEFAULT_HYPERNEAT_WEIGHT_RANGE

public static final double DEFAULT_HYPERNEAT_WEIGHT_RANGE

Default link weight range for HyperNEAT networks.

See Also:

Constant Field Values

Constructor Detail

NEATPopulation

public NEATPopulation()

An empty constructor for serialization.

NEATPopulation

public NEATPopulation(int inputCount,
                      int outputCount,
                      int populationSize)

Construct a starting NEAT population. This does not generate the initial random population of genomes.

Parameters:

inputCount - The input neuron count.

outputCount - The output neuron count.

populationSize - The population size.

NEATPopulation

public NEATPopulation(Substrate theSubstrate,
                      int populationSize)

Construct a starting HyperNEAT population. This does not generate the initial random population of genomes.

Parameters:

theSubstrate - The substrate ID.

populationSize - The population size.

Method Detail

clampWeight

public static double clampWeight(double w,
                                 double weightRange)

Change the weight, do not allow the weight to go out of the weight range.

Parameters:

w - The amount to change the weight by.

weightRange - Specify the weight range. The range is from -weightRange to +weightRange.

Returns:

The new weight value.

assignGeneID

public long assignGeneID()

Returns:

A newly generated gene id.

assignInnovationID

public long assignInnovationID()

Returns:

A newly generated innovation id.

calculateError

public double calculateError(MLDataSet data)

Calculate the error of the ML method, given a dataset.

Specified by:

calculateError in interface MLError

Parameters:

data - The dataset.

Returns:

The error.

compute

public MLData compute(MLData input)

Compute regression.

Specified by:

compute in interface MLRegression

Parameters:

input - The input data.

Returns:

The output data.

getActivationCycles

public int getActivationCycles()

Returns:

Get the activation cycles.

getActivationFunctions

public ChooseObject<ActivationFunction> getActivationFunctions()

Returns:

the activationFunctions

getCODEC

public GeneticCODEC getCODEC()

Returns:

the codec

getGeneIDGenerate

public GenerateID getGeneIDGenerate()

Returns:

the geneIDGenerate

getGenomeFactory

public NEATGenomeFactory getGenomeFactory()

Specified by:

getGenomeFactory in interface Population

Overrides:

getGenomeFactory in class BasicPopulation

Returns:

A factory used to create genomes.

getInitialConnectionDensity

public double getInitialConnectionDensity()

Returns:

the initialConnectionDensity

getInnovationIDGenerate

public GenerateID getInnovationIDGenerate()

Returns:

the innovationIDGenerate

getInnovations

public NEATInnovationList getInnovations()

Returns:

Get the innovations.

getInputCount

public int getInputCount()

Specified by:

getInputCount in interface MLInput

Returns:

The input.

getOutputCount

public int getOutputCount()

Specified by:

getOutputCount in interface MLOutput

Returns:

The output count.

getRandomNumberFactory

public RandomFactory getRandomNumberFactory()

Returns:

the randomNumberFactory

getSubstrate

public Substrate getSubstrate()

Returns:

Returns the hyper-neat substrate.

getSurvivalRate

public double getSurvivalRate()

Returns:

The survival rate, this is the number of genomes used to mate.

getWeightRange

public double getWeightRange()

Returns:

the weightRange

isHyperNEAT

public boolean isHyperNEAT()

Returns:

Returns true if this is a hyperneat population.

reset

public void reset()

Create an initial random population.

setActivationCycles

public void setActivationCycles(int activationCycles)

Set the number of activation cycles to use.

Parameters:

activationCycles - The number of activatino cycles to use.

setCODEC

public void setCODEC(GeneticCODEC codec)

Parameters:

codec - the codec to set

setInitialConnectionDensity

public void setInitialConnectionDensity(double initialConnectionDensity)

Parameters:

initialConnectionDensity - the initialConnectionDensity to set

setInnovations

public void setInnovations(NEATInnovationList theInnovations)

Set the innovation list to use.

Parameters:

theInnovations - The innovation list to use.

setInputCount

public void setInputCount(int inputCount)

Parameters:

inputCount - the inputCount to set

setNEATActivationFunction

public void setNEATActivationFunction(ActivationFunction af)

Specify to use a single activation function. This is typically the case for NEAT, but not for HyperNEAT.

Parameters:

af - The activation function to use.

setOutputCount

public void setOutputCount(int outputCount)

Parameters:

outputCount - the outputCount to set

setRandomNumberFactory

public void setRandomNumberFactory(RandomFactory randomNumberFactory)

Parameters:

randomNumberFactory - the randomNumberFactory to set

setSubstrate

public void setSubstrate(Substrate substrate)

Parameters:

substrate - the substrate to set

setSurvivalRate

public void setSurvivalRate(double theSurvivalRate)

Set the survival rate, this is the percent of the population allowed to mate.

Parameters:

theSurvivalRate - The survival rate.

setWeightRange

public void setWeightRange(double weightRange)

Sets the weight range for links in genomes in this population.

Parameters:

weightRange - The link weight range.