search

neuroevolution-ai / NeuroEvolution-CTRNN_new

MIT License
3 stars 3 forks source link

hof based on validation runs #78

Open bjuergens opened 3 years ago

bjuergens commented 3 years ago

blocked until #70 is merged

maybe blocked by #69

It should be possible to create a HOF based on the performance in validation runs, instead of performance in regular training runs (where luck plays a significant larger role).

The goal is that the HOF will contain much stronger individuals at the end of the experiment, and we can analyze their behavior better

DanielZim commented 3 years ago

This is super important in my opinion to get a meaningful hof at all.