PPLS/D: Parallel Pareto Local Search Based on Decomposition

Abstract

Pareto Local Search (PLS) is a basic building block in many metaheuristics for Multiobjective Combinatorial Optimization Problem (MCOP). In this paper, an enhanced PLS variant called Parallel Pareto Local Search based on Decomposition (PPLS/D) is proposed. PPLS/D improves the efficiency of PLS using the techniques of parallel computation and problem decomposition. It decomposes the original search space into L subregions and executes L parallel processes searching in these subregions simultaneously. Inside each subregion, the PPLS/D process is guided by a unique scalar objective function. PPLS/D differs from the well-known Two Phase Pareto Local Search (2PPLS) in that it uses the scalar objective function to guide every move of the PLS procedure in a fine-grained manner. In the experimental studies, PPLS/D is compared against the basic PLS and a recently proposed PLS variant on the multiobjective Unconstrained Binary Quadratic Programming problems (mUBQPs) and the multiobjective Traveling Salesman Problems (mTSPs) with at most four objectives. The experimental results show that, no matter whether the initial solutions are randomly generated or generated by heuristic methods, PPLS/D always performs significantly better than the other two PLS variants.Jialong Shi is with the

A Hybrid Leader Selection Strategy for Many-Objective Particle Swarm Optimization

“…MaOPs. From among them, the multi-objective evolutionary algorithm based on decomposition (MOEA/D) [11], [12] is the most popular. MOEA/D decomposes a MOP into a set of single-objective sub-problems by using a scalarizing function and these sub-problems are then simultaneously optimized.…”

Parallel pareto local search revisited

“…The rest of this paper is organized as follow. For the paper to be self-contained, we recall in Section 2 the initial design of PPLS as described in [10,11]. In Section 3, we discuss the rst issue of PPLS related to load-imbalance when dealing with an unknown Pareto front shape.…”

“…M = 2, for which PLS is recognized as a state-of-the-art algorithm [6]. Similar to [6,10,11], the neighborhood structure N used in PLS is the 1-bit-ip.…”

“…On the other hand, in some recent papers [10,11], it is shown how to incorporate decomposition into PLS with the aim of bringing parallelism into the scene to enhance search performance. Inspired by the MOEA/D algorithm [12], Shi et al [10,11] proposed to use dierent scalar functions to guide a number of independent PLS processes in parallel towards dierent regions of the objective space. The archives obtained by dierent processes are merged together into one output archive at the end of the run.…”

“…It is, in particular, shown that this high-level parallelizing approach allows one to speedup very substantially the computational ow, while maintaining seemingly the same approximation quality than the original PLS. To the best of our knowledge, the so-obtained Parallel Pareto Local Search (PPLS) [10,11] can in fact be considered as a state-of-the-art high-level parallel design of basic PLS. However, we argue that the mechanism of PPLS still requires further improvement.…”