Recently, Kai Zhang’s team made new progress in the field of evolutionary transfer optimization algorithms. The related research findings were published in IEEE Transactions on Evolutionary Computation in a paper titled “Multi-Task Surrogate-Assisted Search With Bayesian Competitive Knowledge Transfer for Expensive Optimization”.
Innovation: To address the problem of “when to transfer,” existing knowledge transfer methods mostly rely on prior similarity assessment or posterior performance observation. However, in expensive optimization problems with limited evaluation budgets, the former may misjudge task relatedness, while the latter often fails to obtain sufficient evidence in time, thereby leading to negative transfer. To tackle this issue, this study proposes a Bayesian competitive knowledge transfer method (BCKT), which integrates prior beliefs and empirical observations from a Bayesian perspective to dynamically estimate inter-task transferability. By introducing a competition mechanism between optimized solutions from the target task and transferred solutions from source tasks, the proposed method adaptively exploits beneficial knowledge while suppressing negative transfer. The results show that BCKT can effectively improve the performance of multi-task surrogate-assisted search algorithms on both multi-task and many-task expensive optimization problems, and demonstrates good applicability in real-world optimization scenarios.
Abstract: Expensive optimization problems (EOPs) present significant challenges for traditional evolutionary optimization due to their limited evaluation calls. Although surrogate-assisted search (SAS) has become a popular paradigm for addressing EOPs, it still suffers from the cold-start issue. In response to this challenge, knowledge transfer has been gaining popularity for its ability to leverage search experience from potentially related instances, ultimately facilitating head-start optimization for more efficient decision-making. However, the curse of negative transfer persists when applying knowledge transfer to EOPs, primarily due to the inherent limitations of existing methods in assessing knowledge transferability. On the one hand, a priori transferability assessment criteria are intrinsically inaccurate due to their imprecise understandings. On the other hand, a posteriori methods often necessitate sufficient observations to make correct inferences, rendering them inefficient when applied to EOPs. Considering the above, this paper introduces a Bayesian competitive knowledge transfer (BCKT) method developed to improve multi-task SAS (MSAS) when addressing multiple EOPs simultaneously. Specifically, the transferability of knowledge is estimated from a Bayesian perspective that accommodates both prior beliefs and empirical evidence, enabling accurate competition between inner-task and inter-task solutions, ultimately leading to the adaptive use of promising solutions while effectively suppressing inferior ones. The effectiveness of our method in boosting various SAS algorithms for both multi-task and many-task problems is empirically validated, complemented by comparative studies that demonstrate its superiority over peer algorithms and its applicability to real-world scenarios.
IEEE Transactions on Evolutionary Computation is an authoritative journal in the field of evolutionary computation published by the IEEE Computational Intelligence Society. It covers evolutionary computation and related areas, including nature-inspired algorithms, swarm intelligence and population-based optimization methods, genetic algorithms, evolutionary optimization, multi-objective optimization, surrogate-assisted optimization, machine learning and intelligent system design, image processing and machine vision, pattern recognition, evolutionary neural computation, evolutionary fuzzy systems, robotics and control, mathematical modeling, and engineering applications.
The journal was launched in 1997 and is published bimonthly. Its 2024 Impact Factor is 12.0, its 5-year Impact Factor is 14.5, and its 2024 CiteScore is 23.5. It is ranked in JCR Q1 and classified as a Tier 1 journal in the Computer Science category by the Chinese Academy of Sciences. It is also ranked Tier 1 in the subcategories of Computer Science, Artificial Intelligence and Computer Science, Theory & Methods, and is recognized as a Top journal.
Paper link:
https://ieeexplore.ieee.org/document/11300851
Citation:
Lu Y, Zhang K, Xue X, et al. Multi-Task Surrogate-Assisted Search with Bayesian Competitive Knowledge Transfer for Expensive Optimization[J]. IEEE Transactions on Evolutionary Computation, 2025.
