点击次数：

论文类型：会议论文

发表时间：2021-09-11

卷号：2019-October

摘要：Cooperative co-evolution (CC) is one of the effective frameworks to tackle large-scale global optimization (LSGO) problems. For CC framework, problem decomposition is an essential part which can decompose a problem into several smaller sub-problems. However, most decomposition methods just conduct a rough decomposition and some sub-problems still have high dimensionality. Furthermore, CC framework generally has to combine other techniques not just decomposition methods in order to improve efficiency. In this paper, we propose a multi-stage CC method (MSCC) which is divided into three stages: a preliminary stage, a decomposition stage and an optimization stage. First, a self-adaptive scan (SAS) method is used to scan each decision variable and get a good starting point. Second, MSCC method uses a relative contribution-based decomposition (RCD) method to further decompose high dimensional sub-problems. Third, a cumulative resource allocation (CRA) method is performed to self-adaptively select sub-components for optimizing. Finally, we conduct experiments on CEC��2013 LSGO problems to evaluate the efficacy of the MSCC method. When MSCC method combined with CMAES optimizer, it can generate competitive results compared to several state-of-the-art algorithms. ? 2019, Computers and Industrial Engineering. All rights reserved.