Surrogate-assisted differential evolution: A survey

Abstract

Expensive Optimization Problems (EOPs) are a pressing challenge in real-world applications because they require high-quality solutions under tight computational budgets. To tackle this, numerous Surrogate-Assisted Evolutionary Algorithms (SAEAs) have been proposed that combine Evolutionary Algorithms (EAs) with surrogate models. Recently, researchers have conducted systematic surveys on SAEAs to better showcase their potential in solving EOPs. However, most of these efforts have focused on surrogate models, while largely overlooking EAs. This imbalance poses a challenge to the long-term development of SAEAs. Among various SAEAs, Surrogate-Assisted Differential Evolution (SADE) is widely favored by researchers due to the competitive performance of DE in Evolutionary Computation. It has been broadly applied across diverse engineering and scientific domains. Nevertheless, there is still no work that systematically investigates the progress of SADE. To balance the research direction of SAEAs and fill the gap, this paper provides a comprehensive survey of SADE. Its contributions are summarized as follows: This paper first introduces the general optimization framework of SAEAs and briefly reviews the research directions and advances of its key components. Next, a comprehensive survey of SADE is conducted, covering commonly used surrogate models and DE algorithms. It also examines how existing SADE algorithms use DE, performance evaluation methods, and real-world applications. Finally, future challenges and potential research directions are discussed. We hope this work will draw attention to EAs and inspire further research to advance related fields.