Functional-Specialization Multi-Objective Real-Coded Genetic Algorithm: FS-MOGA

Naoki Hamada, Jun Sakuma, Shigenobu Kobayashi, Isao Ono

Abstract: This paper presents a Genetic Algorithm (GA) for multi-objective function optimization. In multi-objective function optimization, we believe that GA should adaptively switch search strategies in the early stage and the last stage for effective search. Non-biased sampling and family-wise alternation are suitable to overcome local Pareto optima in the early stage of search, and extrapolation-directed sampling and population-wise alternation are effective to cover the Pareto front in the last stage. These situation-dependent requests make it difficult to keep good performance through the whole search process by repeating a single strategy. We propose a new GA that switches two search strategies, each of which is specialized for global and local search, respectively. This is done by utilizing the ratio of non-dominated solutions in the population. We examine the effectiveness of the proposed method using benchmarks and a real-world problem.

External IDs:dblp:conf/ppsn/HamadaSKO08