.png){kind=icon}
Return
An evolutionary algorithm based on learning strategies and a predictive model for constrained optimization problems
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2025.102208.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Solving constrained optimization problems (COPs) with evolutionary algorithms is highly active in the evolutionary computation community. Combining evolutionary algorithms with learning strategies is an efficient way to improve the performance for these algorithms. Moreover, according to the present research, constraint handling is also sensitive to the searching results of constrained optimization algorithms. Based on these considerations, this manuscript proposes an evolutionary algorithm assisted by learning strategies and a predictive mode (EALSPM) for solving COPs. Firstly, in order to reduce the constraint pressure and fully utilize the complementary information among different constraints, a classification-collaboration constraint handling technique is designed, where the constraints of the original problem are randomly classified into K classes, then the original problem is decomposed into K subproblems. Correspondingly, K subpopulations are generated, each subpopulation corresponding to a subproblem. Secondly, the evolutionary process is divided into two stages, random learning and directed learning stages. In both stages, these subpopulations interact with each other by using the random learning and directed learning strategies, respectively. As a result, some potentially better solutions can be generated for the original problem. In addition, an improved continuous domain estimation of distribution model is presented based on the information of good individuals and used to predict offspring in evolution. Finally, extensive experiments, on two sets of benchmark test functions from CEC2010 and CEC2017 as well as two practical problems, demonstrate the competitive performance of the proposed method against other state-of-the-art methods.
Journal
IF:
8.5
Papers:
2.2K
Citations:
1.0W
Organization
Cited Papers
No cited papers available