Does SharpNEAT expect (or even require) a static training set? Looking
at the code in NeatEvolutionAlgorithm and this post (http://
groups.google.com/group/sharpneat/browse_thread/thread/
285d5a94db78b7b2) I believe so. But I thought I'd ask before
investigating a work around.
If a static training set is expected/required, how have folks dealt
with either of these problems:
* Curve fitting
* Coevolution (when the only training set data come from head-to-head
matchups of the other genomes)
-Scott