Open oneturkmen opened 5 years ago
Huh, the problem with forcing the disabled genes was that I DID NOT SORT IN DECREASING ORDER! It sorted in an increasing and was instead taking all the worst genes :man_facepalming:
Deep-copying objects from the list (e.g. deepcopy.copy(gene[0]) instead of gene[0]) seemed to have solved the crossover operation problem. Now, the algorithm performs quite well and manages to converge in under 150 generations.
Good news
/test
folder. The documentation is also done.Not so good news
Although I have implemented 99% of the NEAT algorithm, I encountered the following obscurities:
Future ideas