Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
A Novel Niche-based Decomposition Evolutionary Algorithm for Many-objective Optimization
1School of Sciences, Xi'an Technological University, China.
2School of Computer Science and Technology, Xidian University, China.
ABSTRACT
The core idea of the decomposition-based method is dividing multi-objective optimization problems (MOPs) into several subproblems through a set of uniformly distributed weight vectors and optimize them in a cooperative manner. In order to balance population convergence and diversity in existing decomposition-based multi-objective evolutionary algorithms (MOEAs) better, a novel niche-based decomposition evolutionary algorithm for many-objective optimization (MOEA/DD-NICHE) is proposed. MOEA/DD-NICHE is based on MOEA/DD, which combines dominance-based method and decomposition-based method. In the stage of updating population, a new niche diversity maintenance technique is proposed to identify the most crowded subregion and to select the worst solution. It will help to effectively eliminate solutions with similar search directions and enable solutions located in the boundary subregions to survive, thus enhancing the diversity of population. Experimental studies show that MOEA/DD-NICHE has better comprehensive performance than three decomposition-based comparison algorithms on various test problems, especially when dealing with many-objective optimization problems.
Keywords: Evolutionary algorithm, Many-objective, Decomposition, Dominance, Niche, Crowding degree.
Download PDF
1School of Sciences, Xi'an Technological University, China.
2School of Computer Science and Technology, Xidian University, China.
ABSTRACT
The core idea of the decomposition-based method is dividing multi-objective optimization problems (MOPs) into several subproblems through a set of uniformly distributed weight vectors and optimize them in a cooperative manner. In order to balance population convergence and diversity in existing decomposition-based multi-objective evolutionary algorithms (MOEAs) better, a novel niche-based decomposition evolutionary algorithm for many-objective optimization (MOEA/DD-NICHE) is proposed. MOEA/DD-NICHE is based on MOEA/DD, which combines dominance-based method and decomposition-based method. In the stage of updating population, a new niche diversity maintenance technique is proposed to identify the most crowded subregion and to select the worst solution. It will help to effectively eliminate solutions with similar search directions and enable solutions located in the boundary subregions to survive, thus enhancing the diversity of population. Experimental studies show that MOEA/DD-NICHE has better comprehensive performance than three decomposition-based comparison algorithms on various test problems, especially when dealing with many-objective optimization problems.
Keywords: Evolutionary algorithm, Many-objective, Decomposition, Dominance, Niche, Crowding degree.
Download PDF