A host-parasite genetic algorithm for asymmetric tasks

摘要

We present a formalisation of host-parasite coevolution in Evolutionary Computation [2]. The aim is to gain a better understanding of host-parasite Genetic Algorithms (GAs) [3]. We discuss Rosin's [10] competetive theory of games, and show how it relates to host-parasite GAs. We then propose a new host-parasite optimisation algorithm b on this formalisation. The new algorithm takes into account the asymmetry of the two tasks: evolving hosts and evolving parasites. By self-adaptation the algorithm can find a suitable balance between the amount of resources spent on these two tasks. Our results show that this makes it possible to evolve optimal solutions by testing fewer candidates.