Multi-Objective Optimization In Theory and Practice II: Metaheuristic Algorithms

Multi-Objective Optimization In Theory and Practice II: Metaheuristic Algorithms

Multi-Objective Optimization in Theory and Practice is a simplified two-part approach to multi-objective optimization (MOO) problems. This second part focuses on the use of metaheuristic ...
[view complete introduction]

US $
15

*(Excluding Mailing and Handling)



Decomposition-Based and Hybrid Evolutionary Algorithms

Pp. 173-196 (24)

Andre A. Keller

Abstract

Solving an MOO problem may use two different ways. One way consists of decomposing the initial problem into a sequence of subproblems. Each subproblem is solved using information from its neighboring subproblems. We solve all the subproblems simultaneously. A multi-objective evolutionary algorithm based on decomposition (MOEA/D) applies this approach to the Kursawe’s test function. There -lattice are used to approximate convex and nonconvex Pareto-optimal fronts. The metaheuristics or heuristics combine. The incorporation of local search heuristics into an MOEA illustrates this aspect. In the literature, we find how hybridization can be designed. There are three ways to hybridize metaheuristics and heuristics. A first method is to use one algorithm and improve it with other techniques. A second method is to use multiple operators in an EA, and a third method is to develop MOGA solutions by implementing efficient local search. Numerous examples of hybridization exist in the literature. The algorithm M-PAES combines the local search strategy in the Pareto archived evolution strategy (PAES) with the use of GA. The main algorithm PSO can be combined with a local and a global search algorithm.

Affiliation:

Center for Research in Computer Science Signal and Automatic Control of Lille University of Lille – CNRS France.