Extremal Optimisation (EO) is a fairly new entrant into the realms of stochastic based optimisation techniques. Its behaviour differs from other more common algorithms as it alters a poorly performing part of the one solution used without regard to the effect this will have on the quality of the solution. While this means that its performance on assignment problems may be poor if used on its own, this same ‘failing’ makes it a very suitable base for a meta-heuristic. An analysis of the performance of naive EO on the classic bin packing problem is performed in this paper. Results are also presented that show that the same naive EO can be used in a meta-heuristic that performs very well.
Available at: http://works.bepress.com/marcus_randall/11/