Co-evolutionary algorithms: A useful computational abstraction?

Abstract

Interest in co-evolutionary algorithms was triggered in part with Hillis 1991 paper describing his success in using one to evolve sorting networks. Since then there have been heightened expectations for using this nature-inspired technique to improve on the range and power of evolutionary algorithms for solving difficult computation problems. However, after more than two decades of exploring this promise, the results have been somewhat mixed. In this talk I summarize the progress made and the lessons learned with a goal of understanding how they are best used and identify a variety of interesting open issues that need to be explored in order to make further progress in this area.