Does NEAT solve the competing conventions problem?
First let me clarify, as Darrell Whitley pointed out, the competing conventions problem is the problem of having the same functionality represented by different permutations of the same set of connection weights. Thus, NEAT is addressing a broader problem, which really should be termed the “variable length genome problem,” which arises from having different topologies and weight configurations in the same population. That said, the variable length genome problem is the problem of crossing over or comparing neural networks (or other structured phenotypes) with different topologies and weights that implement similar functionalities. NEAT’s method of matching up genomes and measuring compatility for speciation using historical markings may be considered a “solution” to the problem of variable length genomes in the sense that any two genomes can now be matched up in the most sensible manner possible no matter how divergent their topologies, and NEAT can also be considered a solution becaus
