Allen W. Smith, Ph.D.

JSON would be more portable to non-Python implementations, I would think.

OTOH, the `json` module in Python is rather annoying compared to pickle in extending what it can save, from what I can see - you can't specify with a class...

Well, it depends on whether you're talking about initialization of a FloatAttribute (weights, biases, various ones for multiparameter functions) or subsequent mutations. The first can be either using a gaussian/normal...

Interesting! I can see the idea for continuous attributes, I think - it would be, as well as current mutation by (in essence) brownian motion (the addition of a zero-mean...

I have, incidentally, been considering whether it would be worth it to make uniform vs gaussian initialization an evolvable parameter; it would need to convert to something with the same...

One could also do this by evolving a continuous parameter for the chance, each time an initialization or mutation happened, it would use a gaussian or uniform distribution (with equal...

Quite welcome. If the population as a whole had fluctuations in mutations, that could be helpful - alternating between structural and non-structural, for instance; control of "bloat" via alternating between...

By "guess", do you mean "gauss"? I'm not sure whether moving functions into genome_config would be best, or whether checking the genome config for which function to use would be...

BTW, re evolution being too conservative on mutation parameters if allowed to be - see [Natural Selection Fails to Optimize Mutation Rates for Long-Term Adaptation on Rugged Fitness Landscapes](https://doi.org/10.1371/journal.pcbi.1000187). Similarly...

Quite welcome; I like pointing people toward useful information. In a fractured decision space, a rugged - rapidly changing from "place" to "place" - fitness landscape should result if you...