A Novel Objective Reduction Algorithm Using Objective Sampling for Many-Objective Optimization Problems
Abstract— In the field of science and engineering, many problems are Many-objective Optimization Problems (MaOPs), which have more than three objectives. The main difficulty of MaOPs is the true Pareto front is hard to get due to the low selection pressure. However, for some MaOPs, we can reduce the number of objectives to get the non-redundant objectives. In this paper, a novel fast objective reduction algorithm is proposed. Different from other objective reduction algorithms, this algorithm uses a sampling method to get the relationships between objectives by calculating objectives’ improvements. Then, a fast procedure is used to omit the redundant objectives. Finally, experiments show that the proposed algorithm is effective.
Index Terms— many-objective optimization, objective reduction, soft computing, intelligence computation
Minghan Li, Jingxuan Wei, Long Zhao
Department of Computer Science, Xidian University, CHINA
Cite: Minghan Li, Jingxuan Wei, Long Zhao, "A Novel Objective Reduction Algorithm Using Objective Sampling for Many-Objective Optimization Problems," Proceedings of 2018 the 8th International Workshop on Computer Science and Engineering, pp. 713-717, Bangkok, 28-30 June, 2018.