Coevolution, Memory and Balance

Abstract

This paper studies the role of two mechanisms - memory and balance - to exploit the arms race resulting from predator-prey interactions when solving a given problem. Memory ensures that individuals are not only well adapted to the current members of the opposite population but also to earlier generations of opponents. A balanced (co)evolution, on the other hand, adapts the speed of evolution (i.e. the reproduction rate) to the performance of a population. It leads to a steady progress in both populations. Indirectly, a balanced (co)evolution avoids a premature loss of genetic diversity. This in turn, diminishes the need for a long memory span. The current paper shows how both mechanisms can be incorporated in Coevolutionary Genetic Algorithms (CGAs). Empirical results support the importance of, and interaction between, both mechanisms.