Hands-On Neuroevolution with Python

Crossover with an innovation number

There is a piece of unexploited information in the evolutionary process that tells us exactly which genes to match between the genomes of any organism in the topologically diverse population. This is where each gene tells us which ancestor that gene was derived from. The connection genes with the same historical origin represent the same structure, despite possibly having different connection weight values. The historical origins of genes in the NEAT algorithm are represented by incrementally assigned innovation numbers, which allow us to track the chronology of structural mutations.

At the same time, during the crossover, the offspring inherit the innovation numbers of genes from parent genomes. Thus, ...