.png){kind=icon}
.png){kind=icon}
Return
A dual model-based evolutionary framework for dynamic large-scale sparse multiobjective optimization
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2025.102011.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Recently, there has been a growing interest in dynamic multiobjective optimization problems (DMOPs). Although some evolutionary algorithms have been tailored for DMOPs, their effectiveness is limited when handling large-scale DMOPs, especially those characterized by sparsity, where most variables in Pareto-optimal solutions are equal to zero. To address this issue, this paper proposes a dual model-based evolutionary framework to solve dynamic large-scale sparse multiobjective optimization problems (DSMOPs). Specifically, the proposed framework addresses dynamic changes by predicting a new initial population for a static multiobjective optimization evolutionary algorithm in the new environment. Based on the idea of initial population prediction, the proposed framework transforms the large-scale variable prediction into the small-scale variable prediction, where support vector regression is introduced to predict the sparse distributions of the new initial population to reduce the decision space, and multilayer perceptron is performed on the reduced space to predict its continuous distributions. By integrating the two simplified predictions, a two-layer change response mechanism is constructed to ensure both the sparsity and quality of the obtained solutions. In addition, this paper designs the benchmark and real-world test problems to assess the performance of the proposed framework for tackling large-scale DMOPs. Experimental results demonstrate the superiority of the proposed framework compared with the six state-of-the-art algorithms.
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W