A Hybrid Knowledge Transfer for Multitask Optimization

Hai-Xiang Wang¹, Chu-Xiang Li², Zi-Jia Wang^2,*
1 School of Information Engineering, Kaifeng University, Kaifeng, China
2 School of Computer Science and Cyber Engineering, Guangzhou University, Guangzhou, China
* Corresponding Author: Zi-Jia Wang. Email: email
(This article belongs to the Special Issue: Advances in Computational Intelligence for Complex Systems)

Computer Modeling in Engineering & Sciences https://doi.org/10.32604/cmes.2026.086077

Received 23 May 2026; Accepted 17 August 2026; Published online 31 August 2026

Abstract

Evolutionary multitasking optimization (EMTO) is an emerging research direction in evolutionary computation (EC), with its core objective being the collaborative solution of multiple problems through inter-task knowledge transfer (KT). In classical EMTO algorithms, KT typically relies on the direct exchange or crossover of individuals between populations. However, such transfer strategies often follow singular rules or direct transplantation, which struggle to adequately adapt to the dynamically evolving distributional differences between tasks, and may lead to inefficient transfer or even negative transfer. To tackle this issue, this study presents HKTMTO, a multitask differential evolution algorithm built upon two complementary transfer patterns. The algorithm combines estimation of distribution transfer and adaptive elite transfer to construct a hybrid knowledge transfer framework, overcoming the limitations of traditional single individual-level transfer and improving transfer efficiency without triggering negative transfer. Extensive experiments on the CEC2017 multitask benchmark demonstrate the competitive performance of HKTMTO against mainstream EMTO methods.