evolvingfridge

What are possible issues running mpi/ipengines on each processes with 3 nodes and 32 cores on each node ?

@stark7 Unless you hit some precision issues, why not just normalizing all input/output data so it is between 0 and 1 ?

May be I am still missing something but then inputs would look like that: {0/55, 1/140} another words input_A = a/max(a_inputs) input_B = a/max(b_inputs), this way nothing modified just normalized.

Normalizing/Denormalizing Example: Normalizing input: node_input = (input - min(inputs)) / (max(inputs) - min(inputs)) Denormalizing output: node_output = min_bound + node_output*(max_bound - min_bound) Above example is most basic one, there are...

I think its kind of a separate project in by itself pre/post processing can be pretty complex and in my experience always specific to a problem. Will add normalizing example...

I meant instead of random() to use various stochastic process functions, like in finance Random Walk functions and in physics Brownian Motion, random() can be implemented with jump diffusion. This...

I was thinking individual genome would have different mutation rates, some genome can be more aggressive in updating attributes and others would be more passive (what changes is to). Right...

Thanks for pointing out overly-conservative mutation rates, I did not thing about that, but don't see as of now how this would happen if jump diffusion will be used, because...

Just to make sure I understand it correctly, from out conversation it is good idea to move guess, random() and possibly choice to genome_config initially, before implementing stagnation functions.

Yes, sorry about miss spelling, I meant it should be implemented similarly as in [multiparam](https://github.com/drallensmith/neat-python/tree/multiparam_funcs/neat) fork. It looks like, if there is no knowledge of past then Gaussian distribution is...