Joined: 18 Nov 07
I read through your presentation - it was interesting to see how different search methods compare.
I was wondering about something you didn't mention: in nature, evolution doesn't care if a mutation does no good or even if it harms the individual in some way, to a limited extent. This is especially important in group societies, where detrimental mutations won't necessarily harm the survival chances of the group.
By allowing such mutations, you can at times get interesting new developments: by going off on an apparently meaningless or even detrimental side path, startling new developments can happen which are either
1) an unforseen advantage of a combination of earlier changes, or are
2) the result of compensating for a detrimental change which is later (randomly) undone.
As far as I could tell from the slides, you do not allow detrimental mutations to happen - only ones that make no difference either way. Have you looked into whether allowing such mutations to propagate for a limited amount of generations could be helpful to the searches?
©2019 Astroinformatics Group