Co-evolving memetic algorithms: Initial investigations

This paper presents and examines the behaviour of a system whereby the rules governing local search within a Memetic Algorithm are co-evolved alongside the problem representation. We describe the rationale for such a system, and the implementation of a simple version in which the evolving rules are encoded as (condition:action) patterns applied to the problem representation, and are effectively self-adapted. We investigate the behaviour of the algorithm on a test suite of problems, and show significant performance improvements over a simple Genetic Algorithm, a Memetic Algorithm using a fixed neighbourhood function, and a similar Memetic Algorithm which uses random rules, i.e. with the learning mechanism disabled. Analysis of these results enables us to draw some conclusions about the way that even the simplified system is able to discover and exploit different forms of structure and regularities within the problems. We suggest that this “meta-learning” of problem features provides a means both of creating highly scaleable algorithms, and of capturing features of the solution space in an understandable form