The structure of a population can dramatically alter the adaptive behaviour of an evolving system. Researchers of evolutionary computation model this concept of geographical separation through the use of spatially-structured evolutionary algorithms (SSEAs); individuals are placed at nodes in a graph and selection/recombination is restricted to nodes that share common edges. Previous work claims that SSEAs are able to support the discovery of multiple solutions to a problem within a single run. However, this behaviour is heavily dependent on the nature of the problem and is rarely observed in practice.
In this talk, I will show how niching methods previously used only in unstructured populations may be included into the selection process of SSEAs. The results presented will show that the performance of this hybrid approach can exceed that observed when using an SSEA on its own (or indeed, when using a niching method within an unstructured population).
Last modified: Thursday, 28-Jul-2005 17:23:30 NZST
This page is maintained by the seminar list administrator.