Agent-based Evolutionary and Memetic Black-box Discrete Optimization

https://doi.org/10.1016/j.procs.2017.05.173Get rights and content
Under a Creative Commons license
open access

Abstract

Hybridizing agent-based paradigm with evolutionary or memetic computation can enhance the field of meta-heuristics in a significant way, giving to usually passive individuals autonomy and capabilities of perception and interaction with other ones. In the article, an evolutionary multi-agent system (EMAS) is applied to solve difficult discrete benchmark problems without any domain-specific knowledge—thus they may be called “black-box” ones. As a means for comparison, a parallel evolutionary algorithm (constructed along with Michalewicz model) versus evolutionary and memetic versions of EMAS are used. The obtained results point out that EMAS is significantly more efficient than classical evolutionary algorithms and also finds better results in the examined problem instances.

Keywords

Agent-based computing
metaheuristics
discrete global optimization
memetic algorithms

Cited by (0)