CodeReclaimers / neat-python

Python implementation of the NEAT neuroevolution algorithm
BSD 3-Clause "New" or "Revised" License
1.42k stars 493 forks source link

best fitness stays constant #157

Open Mocha2007 opened 5 years ago

Mocha2007 commented 5 years ago

Best fitness quickly falls to 3 and just sits there in just a few generations. However, the best possible score is 4.

My code and configurations are identical to config-feedforward and evolve-minimal.py, with the sole exception that the element in the last tuple of xor_outputs is 2.0 instead of 0.0.

Why does this not work? I tried looking through the documentation, settings, other examples, etc. By this point I've tried changing most of the values in the config - nothing changes.

What exactly am I missing here?