Package net.bmahe.genetics4j.neat

Class NeatUtils

java.lang.Object

net.bmahe.genetics4j.neat.NeatUtils

public class NeatUtils
extends Object

Utility class providing core algorithmic operations for the NEAT (NeuroEvolution of Augmenting Topologies) algorithm.

NeatUtils contains essential algorithms and helper methods for implementing NEAT neural network evolution, including network topology analysis, compatibility distance calculation, speciation, and structural operations. These utilities support the NEAT algorithm's key features of topology innovation, structural mutation, and species-based population organization.

Key functionality areas:

• Network topology analysis: Computing network layers, forward/backward connections, and dead node detection
• Compatibility distance: Measuring genetic similarity between neural networks for speciation
• Speciation management: Organizing populations into species based on genetic similarity
• Structural analysis: Analyzing network connectivity patterns and structural properties

NEAT algorithm integration:

• Innovation tracking: Support for historical marking and innovation numbers
• Structural mutations: Utilities for add-node and add-connection operations
• Network evaluation: Layer-based network evaluation ordering
• Population diversity: Species-based diversity maintenance

Core NEAT concepts implemented:

• Genetic similarity: Compatibility distance based on excess, disjoint, and weight differences
• Topological innovation: Structural changes tracked through innovation numbers
• Speciation: Dynamic species formation based on genetic distance thresholds
• Network evaluation: Feed-forward evaluation through computed network layers

Algorithmic foundations:

• Graph algorithms: Topological sorting, connectivity analysis, and layer computation
• Genetic distance metrics: NEAT-specific compatibility distance calculation
• Population clustering: Species formation and maintenance algorithms
• Network optimization: Dead node removal and structural simplification

See Also:

• NeatChromosome
• Connection
• Species
• InnovationManager

Constructor Summary

Constructors

Modifier	Constructor	Description
private	NeatUtils()

Method Summary

Modifier and Type	Method	Description
static float	compatibilityDistance(List<Connection> firstConnections, List<Connection> secondConnections, float c1, float c2, float c3)
static float	compatibilityDistance(Genotype genotype1, Genotype genotype2, int chromosomeIndex, float c1, float c2, float c3)
static Map<Integer,Set<Connection>>	computeBackwardConnections(List<Connection> connections)
static Map<Integer,Set<Integer>>	computeBackwardLinks(List<Connection> connections)
static Set<Integer>	computeDeadNodes(List<Connection> connections, Map<Integer,Set<Integer>> forwardConnections, Map<Integer,Set<Integer>> backwardConnections, Set<Integer> outputNodeIndices)	Working backward from the output nodes, we identify the nodes that did not get visited as dead nodes
static Map<Integer,Set<Integer>>	computeForwardLinks(List<Connection> connections)
static List<List<Integer>>	partitionLayersNodes(Set<Integer> inputNodeIndices, Set<Integer> outputNodeIndices, List<Connection> connections)
static <T extends Comparable<T>> List<Species<T>>	speciate(RandomGenerator random, SpeciesIdGenerator speciesIdGenerator, List<Species<T>> seedSpecies, Population<T> population, BiPredicate<Individual<T>,Individual<T>> speciesPredicate)

Methods inherited from class java.lang.Object

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

Constructor Details

NeatUtils

private NeatUtils()

Method Details

computeDeadNodes

public static Set<Integer> computeDeadNodes(List<Connection> connections,
 Map<Integer,Set<Integer>> forwardConnections,
 Map<Integer,Set<Integer>> backwardConnections,
 Set<Integer> outputNodeIndices)

Working backward from the output nodes, we identify the nodes that did not get visited as dead nodes

Parameters:

connections -

forwardConnections -

backwardConnections -

outputNodeIndices -

Returns:

computeForwardLinks

public static Map<Integer,Set<Integer>> computeForwardLinks(List<Connection> connections)

computeBackwardLinks

public static Map<Integer,Set<Integer>> computeBackwardLinks(List<Connection> connections)

computeBackwardConnections

public static Map<Integer,Set<Connection>> computeBackwardConnections(List<Connection> connections)

partitionLayersNodes

public static List<List<Integer>> partitionLayersNodes(Set<Integer> inputNodeIndices,
 Set<Integer> outputNodeIndices,
 List<Connection> connections)

compatibilityDistance

public static float compatibilityDistance(List<Connection> firstConnections,
 List<Connection> secondConnections,
 float c1,
 float c2,
 float c3)

compatibilityDistance

public static float compatibilityDistance(Genotype genotype1,
 Genotype genotype2,
 int chromosomeIndex,
 float c1,
 float c2,
 float c3)

speciate

public static <T extends Comparable<T>>
List<Species<T>> speciate(RandomGenerator random,
 SpeciesIdGenerator speciesIdGenerator,
 List<Species<T>> seedSpecies,
 Population<T> population,
 BiPredicate<Individual<T>,Individual<T>> speciesPredicate)