Abstract

In this paper, we study the effect of Meta-Lamarckian learning on the performance of a generic hybrid Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D) to solve a well-known combinatorial Multi-Objective Optimization (MOO) problem. We study the hybridization of MOEA/D with a pool of six general-purpose heuristics so as to locally optimize the solutions during the evolution. We initially consider the six individualistic hybrid MOEA/D's, in which at every step of the evolution the same local search heuristic from the generic pool is applied. MOEA/D is then enriched with a learning strategy that, based on the problem's properties and objective functions, adaptively selects at each step of the evolution and for each problem neighbourhood the best performing local search heuristic from the generic pool of heuristics. The proposed method is evaluated on various test instances of a multi-objective Permutation Flow Shop Scheduling Problem (MO-PFFSP): given a set of jobs and a series of machines, the corresponding processing time of each job on every machine and the due dates of each job, determine a processing order of the jobs on each machine, so as to simultaneously minimize the makespan (total completion time), and the maximum job tardiness. The results of our experimental studies suggest that the proposed method successfully learns the behaviour of individual local search heuristics during the evolution outperforming in terms of both convergence and diversity the conventional MOEA/D and the individualistic hybrid MOEA/D's. The proposed method does not utilize any problem-specific heuristics, and as a result, is readily applicable to other combinatorial MOO problems.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call