Click here for full text:
Visualising Evolutionary Pathways in Real-World Search Spaces
Layzell, Paul
HPL-2002-308
Keyword(s): neutral networks; neutrality fitness; landscape; EA; GA; genetic
Abstract: A new method of visualising search spaces on a two- dimensional surface is presented, and applied to real search spaces created by the artificial evolution of electronic circuits. The visualisation method is inspired by work on search space neutrality, and its links with evolvability, both of which are discussed. The search space is transformed into a connected graph. Vertices on the graph are connected if their Hamming distance from each other is one, and a clustering algorithm reveals important inter-cluster pathways that can be reached by single mutations.
6 Pages
Back to Index
| 
printable version
