search

geneura-papers / 2017-GPRuleRefinement

Repository for the GPRuleRefinement paper to be sent to a Journal.
Artistic License 2.0
0 stars 0 forks source link

To implement a hierarchical fitness #13

Closed unintendedbear closed 7 years ago

unintendedbear commented 8 years ago

So if two individuals obtain the same fitness, the one which survives is the one with the smalest tree.

JJ commented 8 years ago

That creates pressure towards smaller trees, biasing search. You shouldn't try to do that unless it's really a problem. It's better to cap the tree size at a particular value.

unintendedbear commented 8 years ago

So you propose to repeat the experiments trying different max. depth of the tree?

JJ commented 8 years ago

Specifically, I don't propose anything. I'm just saying that hierarchical fitness functions are Evil. But that would be a good option, yes. Better than hierarchical fitness, anyway.