Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
SPEA2-NIA: A Strength Pareto Evolutionary Algorithm with a Neighborhood Interval Advantage Indicator
1School of Information Engineering, JiangXi University of Science and Technology, China.
2School of Mathematical and Computer Science, Gannan Normal University, China.
ABSTRACT
Significant progress has been made in addressing complex multiobjective optimization problems by utilizing multiobjective evolutionary algorithms. Despite the excellent performance demonstrated by many existing algorithms, challenges related to diversity maintenance and convergence achievement are encountered. These challenges are manifested in the potential loss of crucial solutions or the risk of convergence to local optima, consequently impeding the effective exploration of the complete Pareto front (PF). A novel indicator called the neighborhood interval advantage (NIA) is proposed in this paper. This indicator is integrated into the SPEA2 algorithm, resulting in the development of the SPEA2-NIA algorithm. The NIA indicator considers both the convergence and diversity of individuals within their neighborhoods. Specifically, the strategy involves the division of an individual's neighborhood, followed by calculating the normalized sum of inferiority differences among the solutions within said neighborhood, based on their function values. Through a series of experiments and performance comparisons, it is demonstrated that the good performance of the SPEA2-NIA algorithm, compared to traditional algorithms, is especially pronounced when dealing with complex multiobjective problems.
Keywords: Multiobjective, Optimization, Indicator, Evolutionary algorithms, Neighborhood computation, Diversity.
Download PDF
1School of Information Engineering, JiangXi University of Science and Technology, China.
2School of Mathematical and Computer Science, Gannan Normal University, China.
ABSTRACT
Significant progress has been made in addressing complex multiobjective optimization problems by utilizing multiobjective evolutionary algorithms. Despite the excellent performance demonstrated by many existing algorithms, challenges related to diversity maintenance and convergence achievement are encountered. These challenges are manifested in the potential loss of crucial solutions or the risk of convergence to local optima, consequently impeding the effective exploration of the complete Pareto front (PF). A novel indicator called the neighborhood interval advantage (NIA) is proposed in this paper. This indicator is integrated into the SPEA2 algorithm, resulting in the development of the SPEA2-NIA algorithm. The NIA indicator considers both the convergence and diversity of individuals within their neighborhoods. Specifically, the strategy involves the division of an individual's neighborhood, followed by calculating the normalized sum of inferiority differences among the solutions within said neighborhood, based on their function values. Through a series of experiments and performance comparisons, it is demonstrated that the good performance of the SPEA2-NIA algorithm, compared to traditional algorithms, is especially pronounced when dealing with complex multiobjective problems.
Keywords: Multiobjective, Optimization, Indicator, Evolutionary algorithms, Neighborhood computation, Diversity.
Download PDF