Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
A Surrogate Model Driven Many-Objective Evolutionary Algorithm
School of Mathematics and Statistics, Qinghai Normal University, Xining.
ABSTRACT
It is critical to apply the Pareto dominance relation in multi-objective optimization problems since it can guarantee good convergence and diversity. But the Pareto dominance strategy may lose its effectiveness when there exists many objectives. To solve this problem, a surrogate model driven many-objective evolutionary algorithm is proposed, named MaOEA-SMDA. First of all, a surrogate model strategy is designed to transform a many-objective model into a bi-objective model. Then, the population is used to generate a convergence archive and a diversity archive in the process of evolution to ensure the diversity of population. Experimental studies on multiple benchmark problems show that MaOEA-SMDA is competitive for different number of objectives. Compared with five current state-of-the-art algorithms, MaOEA-SMDA can cope with MaOPs with better performance.
Keywords: Surrogate model technology, Many-objective optimization, Evolutionary algorithm, Dual-archive, Biobjective model.
Download PDF
School of Mathematics and Statistics, Qinghai Normal University, Xining.
ABSTRACT
It is critical to apply the Pareto dominance relation in multi-objective optimization problems since it can guarantee good convergence and diversity. But the Pareto dominance strategy may lose its effectiveness when there exists many objectives. To solve this problem, a surrogate model driven many-objective evolutionary algorithm is proposed, named MaOEA-SMDA. First of all, a surrogate model strategy is designed to transform a many-objective model into a bi-objective model. Then, the population is used to generate a convergence archive and a diversity archive in the process of evolution to ensure the diversity of population. Experimental studies on multiple benchmark problems show that MaOEA-SMDA is competitive for different number of objectives. Compared with five current state-of-the-art algorithms, MaOEA-SMDA can cope with MaOPs with better performance.
Keywords: Surrogate model technology, Many-objective optimization, Evolutionary algorithm, Dual-archive, Biobjective model.
Download PDF