.png){kind=icon}
Return
An efficient decomposition based multiobjective evolutionary algorithm with clustering covariance matrix adaptation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.matcom.2026.08.018.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Recently, covariance matrix adaptation (CMA) has been widely adopted within the design of multiobjective evolutionary algorithms. However, the existing model of CMA is built for each solution in the whole or part of the population, which may lose the global distribution information of the population and result in a relatively low search efficiency. To address this issue, this paper proposes a multiobjective covariance matrix adaptation evolutionary algorithm based on decomposition and clustering, called C-MCMA/D. The key components lie in two aspects: the improving CMA modeling and the adaptive hybrid search mechanism. First, the weight vectors under the decomposition framework are grouped by K-means clustering in the objective space. The CMA is then modeled with all individuals within each group, replacing the modeling method based on a single solution. This achieves the approximation of a more effective region of the Pareto solutions. Second, to fuse global and individual information, an adaptive hybrid mechanism is introduced through the parameter a to dynamically balance the CMA and differential evolution operators, based on the historical information about the evolutionary process. The experimental results demonstrate that C-MCMA/D significantly outperforms the compared algorithms on test instances. It suggests that C-MCMA/D will be a competitive candidate for multiobjective optimization.
Keywords:
Multiobjective optimization
Decomposition approach
Covariance matrix adaptation
Adaptive strategy
Journal
IF:
4.4
Papers:
949
Citations:
1.0W
Organization
Cited Papers
No cited papers available