Abstract | ||
---|---|---|
Multi-objective evolutionary algorithm (MOEA) based on user preference has become a popular research topic in recent years. Target region is a simple form of user preference. In this paper, it is defined as a region consisting of specific interval on each dimension of the objective space. Target region works as guidance in the search process. However, like other forms of user preference, it is difficult to set target region properly without sufficient prior knowledge. To solve this problem, a mechanism to make target region self-adaptive is proposed. It utilizes the information from non-dominated solutions. If there are solutions dominating target region, the target region will be directly updated based on the range of these solutions. Otherwise, target region will keep exploring until stop criterion is satisfied. The mechanism is applied on multi-objective particle swarm optimization (MOPSO). Experimental results demonstrate that MOPSO with self-adaptive target region has better convergence and it avoids the population getting stuck in local optimum compared to MOPSO. |
Year | DOI | Venue |
---|---|---|
2020 | 10.1109/CoDIT49905.2020.9263835 | 2020 7th International Conference on Control, Decision and Information Technologies (CoDIT) |
Keywords | DocType | Volume |
multiobjective evolutionary algorithm,user preference,target region works,multiobjective particle swarm optimization,self-adaptive target region,search process | Conference | 1 |
ISSN | ISBN | Citations |
2576-3547 | 978-1-7281-5954-6 | 0 |
PageRank | References | Authors |
0.34 | 9 | 2 |