Multiple parents, multiple crossovers and incest prevention in evolutionary computation

Abstract

Multimodal optimization is an always present topic in Computer Systems and Networks design and implementation. - Evolutionary Computation is an emergent field which provides new heuristics to function optimization where traditional approaches make the problem computationally intractable. The present contribution gives an insight of the current enhancements that can be done in evolutionary techniques, attempting to balance exploitation and explotation to avoid premature convergence during the search process. Multiple parents, multiple crossovers and incest prevention are three different techniques that when combined showed a substantial benefit: The set of suboptimal solutions are concentrated nearby the optimal solution. This paper shows the design, implementation and partial performance results when a combination of multiple crossovers on multiple parents and incest prevention is applied to an evolutionary algorithm optimizing two difficult multimodal functions.