GET IN TOUCH:

Department of Electronic Engineering, University of York, Heslington, York, YO10 5DD

© 2018 by Julian F. Miller. Proudly created with Wix.com

CONTACT US:

This page contains thoughts and suggestions for future work regarding CGP. If you use any of these ideas I would appreciate you acknowledging them in your paper. I started this "blog" on Nov 2, 2018.

1. Modules (added on Nov 2, 2018)

I think it would be interesting to use the real-valued representations of CGP (either RVCGP or Positional CGP) when using module acquisition. Does it make it easier?

 

2. Mutation (added on Nov 2, 2018)

 

Mutate nodes by randomly choosing a node and then connecting it  to the least active node (over the evolutionary run so far). This should remove the activity bias that exists whereby nodes nearer to inputs tend to be most active and nodes nearest outputs are least active. I think the activity bias is not useful and the new mutation might be much more effective.