Solving large-scale multiobjective optimization via the probabilistic prediction model

Haokai Hong, Kai Ye, Min Jiang, Donglin Cao, Kay Chen Tan

Abstract: The characteristic of large-scale multiobjective optimization problems (LSMOPs) is optimizing multiple conflicting objectives while considering thousands of decision variables at the same time. An efficient optimization algorithm for LSMOPs should have the ability to search a large decision space and find the global optimum in the objective space. Maintaining the diversity of the population is one of the effective ways to locate the Pareto optimal set in a large search space. In this paper, we propose a large-scale multiobjective optimization algorithm based on the probabilistic prediction model, called LMOPPM, to establish a generating-filtering strategy and tackle the LSMOP. The proposed method improves the diversity of the population through importance sampling and enhances the convergence of the population via a trend prediction model. Furthermore, due to the adoption of the individual-based evolutionary mechanism, the computational costs of the proposed method are less relevant to the number of decision variables, thus avoiding high time complexity. We compared the proposed algorithm with several state-of-the-art algorithms on benchmark functions. The experimental results and complexity analysis demonstrate that the proposed algorithm provides significant improvements in terms of performance and computational efficiency in large-scale multiobjective optimization.