A Review on Decomposition-Based Multi-Objectives Evolutionary Algorithms

Abstract

The real-world optimization problems always have many conflicting objectives that need to be optimized at once. A comprehensive number of research studies try to solve such kind of problems. Multi-Objectives Evolutionary-Algorithms using Decomposition (MOEA/D) is one of the most powerful algorithms that solves both multi and many objective optimization problems. The basic idea of such algorithms is to convert the complex Multi-Objective Optimization Problem (MO-OP) into a set of uniobjective subproblems. This conversion is performed with the help of the information acquired from the neighborhood of each subproblem. The algorithm could efficiently solve the tradeoffs between both diversity of the proposed solutions and the convergence of the algorithm. Due to the simplicity and the efficiency of the algorithm, different researches investigated the improvement and adaptation of MOEA/D. In this paper, a review of the different decomposition-based algorithms is proposed. The research studies covered in this paper is categorized into four groups; weight vector generation, scalarization and aggregation strategies, the MOEA/D variants and the MOEA/D real-world applications.