Sexual Recombination in Self-Organizing Interaction Networks

We build on recent advances in the design of self-organizing interaction networks by introducing a sexual variant of an existing asexual, mutation-limited algorithm. Both the asexual and sexual variants are tested on benchmark optimization problems with varying levels of problem difficulty, deception, and epistasis. Specifically, we investigate algorithm performance on Massively Multimodal Deceptive Problems and NK Landscapes. In the former case, we find that sexual recombination improves solution quality for all problem instances considered; in the latter case, sexual recombination is not found to offer any significant improvement. We conclude that sexual recombination in self-organizing interaction networks may improve solution quality in problem domains with deception, and discuss directions for future research.

• Type: Conference
• Year: 2010
• Where: EVOW
• Authors: Joshua L. Payne, Jason H. Moore
• Keywords: Artificial Intelligence, EVOW 2010, Self-organizing Interaction Networks, Sexual Recombination, Sexual Variants