Knowledge transfer in evolutionary multi-task optimization: A survey

Abstract

Evolutionary multi-task optimization (EMTO) is an optimization algorithm designed to optimize multiple tasks simultaneously. In real life, different tasks often correlate to each other, and there exists implicit knowledge or skills common to these tasks. In EMTO, such common knowledge is utilized in an evolutionary optimization process, and EMTO aims to transfer knowledge across different tasks to improve performance in solving each task independently. Therefore, an effective and efficient knowledge transfer mechanism is critical to ensure the success of EMTO. To this end, this survey focuses on the current research progress of knowledge transfer methods in EMTO and proposes a systematic multi-level taxonomy to categorize the existing work. The conduct of categorization is based on two key problems: when and how to perform knowledge transfer. Based on the proposed taxonomy, we also discuss the possibility of integrating various approaches under different categories and applying transfer learning approaches to EMTO. Throughout these discussions, this survey paper aims to identify the potential research directions for improving knowledge transfer performance in EMTO.