Package com.anji.neat

com.anji.neat contains classes implementing NEAT framework, including genes, mutation operators, and fitness function used for XOR.

See:
Description

Class Summary

AddConnectionMutationOperator	Implements NEAT add connection mutation inspired by Evolving Neural Networks through Augmenting Topologies .
AddNeuronMutationOperator	Implements NEAT add node mutation inspired by Evolving Neural Networks through Augmenting Topologies .
ConnectionAllele	Gene corresponding to NEAT connection gene according to Evolving Neural Networks through Augmenting Topologies
ConnectionGene	Gene corresponding to NEAT connection gene according to Evolving Neural Networks through Augmenting Topologies
Evolver	Configures and performs an ANJI evolutionary run.
NeatActivator	Transcribe Chromosome object (loaded from persistence if necessary) into Activator object and activate it with specified stimuli.
NeatChromosomeUtility	Utility class capturing functionality pertaining to NEAT neuron and connection genes.
NeatConfiguration	Extension of JGAP configuration with NEAT-specific features added.
NeatCrossoverReproductionOperator	Implements NEAT crossover reproduction according to Evolving Neural Networks through Augmenting Topologies .
NeatIdMap	Extension of JGAP configuration with NEAT-specific features added.
NeatTargetFitnessFunction	Fitness function where error is subtracted from max fitness, then squared.
NeuronAllele	Gene corresponding to NEAT node gene according to Evolving Neural Networks through Augmenting Topologies
NeuronGene	Gene corresponding to NEAT node gene according to Evolving Neural Networks through Augmenting Topologies
NeuronType	Enumerated type representing flavors of neurons: input, output, hidden.
PruneMutationOperator	Removes neurons and connections that do not affect the activation of the network.
RemoveConnectionMutationOperator	Implements remove connection mutation, using one of 3 strategies: SKEWED - Probability of connection being removed ranges from 0 (if weight magnitude >= max weight removed) to mutation rate (if weight is 0) ALL - All connections have equal likelihood of being removed, regardless of weight.
RemoveConnectionMutationOperator.Strategy	Enumerated type for remove connection strategies.
SingleTopologicalMutationOperator
WeightMagnitudeComparator	sorts ConnectionGenes in ascending order of weight magnitude
WeightMutationOperator	Implements NEAT perturb connection weight mutation according to Evolving Neural Networks through Augmenting Topologies .

Package com.anji.neat Description

com.anji.neat contains classes implementing NEAT framework, including genes, mutation operators, and fitness function used for XOR. Also includes subtractive mutations added to NEAT by James and Tucker for GECCO 2004.