Evolution Algorithm

Singleton class that is responsible for initialization of the evolution algorithm. EvolutionAlgorithm.prefab must be always present in the scene, because it acts as a container for the variables, that are used in other scripts.

classDiagram
    EvolutionAlgorithm <.. WeaponManager : Dependency
    EvolutionAlgorithm <.. AbstractWeapon : Dependency
    EvolutionAlgorithm <.. EvoWeapon : Dependency

---

Evolution Algorithm

Property	Type	Description
PopulationSize	int	Gets or sets the number of genomes to create.
Generation	uint	The current evolution algorithm generation.
InputCount	int	Gets or sets the number of network inputs.
OutputCount	int	Gets or sets the number of network outputs.
CloneOffspringCount	int	Gets or sets the number of networks created by asexual reproduction.
SexualOffspringCount	int	Gets or sets the number of networks created by sexual reproduction.

You can change these properties in InitializeEvolutionAlgorithmParams method:

private void InitializeEvolutionAlgorithmParams() {
    _activationScheme = NetworkActivationScheme.CreateAcyclicScheme();
    PopulationSize = 6; // Modify this parameter to change number of weapons
    CloneOffspringCount = 2;
    SexualOffspringCount = PopulationSize - CloneOffspringCount;
    // Do not change this
    Generation = 0;
    InputCount = 3;
    OutputCount = 5;
}

Method	Description
CreateNewGeneration	Creates a new generation of genomes using the user-selected genomes as parents.
CreateRandomPopulation	Creates a random population of genomes.

Activation Functions

Function	id	Probability
Linear	0	0.1
Sine	1	0.1
ArcTan	2	0.1
BipolarGaussian	3	0.1
BipolarSigmoid	4	0.1
LogisticFunction	5	0.1
QuadraticSigmoid	6	0.1
TanH	7	0.1
ArcSinH	8	0.1
ReLU	9	0.1

Neat Genome Parameters

Represents parameters specific to NEAT genomes. E.g. parameters that describe probabilities for the different types of mutation and the proportion of possible connections to instantiate between input and output neurons within the initial population.

You can change these parameters as you see fit, but FeedfowardOnly must be true.

Property	Type	Description
FeedforwardOnly	bool	Gets or sets a boolean that indicates if NEAT should produce only feed-forward networks (no recurrent/cyclic connection paths).
ConnectionWeightRange	double	Gets or sets the connection weight range to use in NEAT genomes. E.g. a value of 5 defines a weight range of -5 to 5. The weight range is strictly enforced - e.g. when creating new connections and mutating existing ones.
InitialInterconnections Proportion	double	Gets or sets a proportion that specifies the number of interconnections to make between input and output neurons in an initial random population. This is a proportion of the total number of possible interconnections.
DisjointExcessGenes RecombinedProbability	double	Gets or sets the probability that all excess and disjoint genes are copied into an offspring genome during sexual reproduction. Currently, the excess/disjoint genes are copied in an all or nothing strategy.
ConnectionWeight MutationProbability	double	Gets or sets the probability that a genome mutation operates on genome connection weights.
AddNode MutationProbability	double	Gets or sets the probability that a genome mutation is an add node mutation.
AddConnection MutationProbability	double	Gets or sets the probability that a genome mutation is an add connection mutation.
NodeAuxState MutationProbability	double	Gets or sets the probability that a genome mutation is a node auxiliary state mutation.
DeleteConnection MutationProbability	double	Gets or sets the probability that a genome mutation is a delete connection mutation.