(selection) for evolution (e.g. methods.Selection.FITNESSPROPORTIONATE).
Elitism of every evolution loop. Elitism in genetic algorithms.
Calculate scores of every network in population.
There are 3 ways here:
the fitness function provided by the user
the dataset provided by the user
the loss function
Create networks to initialize this population
the template which gets copied for the whole population if provided
if no template is given creating new networks with this input size
if no template is given creating new networks with this output size
Removing a percentage of bad networks from each species
the percentage of kills in each species
Evolves the network to reach a lower error on a dataset using the NEAT algorithm
the evolve options EvolveOptions
Returns the best network.
Returns a random network of this population
Remove bad performing species
If the population stagnates kill all, but the 2 best species
Kills species that continue to stagnate
Logging after one epoch.
Mutate the population.
Sets allowed mutation methods for evolution, a random mutation method will be chosen from the array when mutation occurs. Optional, but default methods are non-recurrent.
Sets the mutation rate. If set to 0.3, 30% of the new population will be mutated.
If mutation occurs (randomNumber < mutationRate), sets amount of times a mutation method will be applied to the network.
more options relevant for the mutation functions
Sort networks in descending order
Sort species by their highest scoring network
Create species. Put networks with a specific distance into the same species
Sum up the average adjusted scores from all species.
Generated using TypeDoc
Breeding the new population