.png){kind=icon}
Return
Evolutionary multitasking optimization based on cross-task association mapping strategy
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.129580.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In multitasking optimization, knowledge transfer between tasks through subspace generation has been widely employed to enhance the convergence performance of algorithms. However, this approach fails to account for the inter-task knowledge mapping relationships. Therefore, cross-task knowledge transfer during the optimization process remains inherently blind, potentially leading to mismatched subspace information and consequently degrading the algorithm’s performance. To address this issue, this paper proposes a multitask evolutionary algorithm based on an association mapping strategy and an adaptive population reuse mechanism, namely PA-MTEA. Specifically, to fully represent the correlations between multitask domains and enhance the adaptability of transfer solutions in target tasks, this paper introduces a subspace projection strategy based on partial least squares, which achieves the correlation mapping between the source and target tasks during the dimensionality reduction of the search space. Additionally, to further enhance knowledge transfer across tasks, an alignment matrix is obtained by adjusting the subspace Bregman divergence after deriving the respective subspaces, minimizing variability between task domains. Finally, to balance the global exploration of algorithms with local exploitation, an adaptive population reuse mechanism based on the residual structure is designed. This mechanism reuses historically successful individuals to guide the evolutionary direction of the population, thus improving the algorithm’s convergence performance. Experimental results on various benchmark suites and real-world cases demonstrate that PA-MTEA exhibits significantly superior performance compared to six other advanced multitask optimization algorithms.
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W
Organization
View more organizations
Cited Papers
No cited papers available