TITLE:
Research on Adaptive Surrogate-Assisted Evolutionary Algorithm Based on SISSO Model Selector
AUTHORS:
Baolei Li, Weizhong Qiu, Chunying Chen, Xiang Liu, Ziwei Zhao, Wentao You
KEYWORDS:
SISSO, Evolutionary Algorithm, Adaptive, Features
JOURNAL NAME:
Journal of Computer and Communications,
Vol.14 No.2,
February
14,
2026
ABSTRACT: Surrogate-assisted evolutionary algorithms are widely used to solve expensive optimization problems due to their high search efficiency. However, a single model struggles to fit various fitness landscapes with different characteristics. How to adaptively select the most suitable model based on the problem’s fitness landscape characteristics remains an open research area. Therefore, this paper proposes an adaptive surrogate model selection framework based on an SISSO model selector. This method extracts problem features from the optimization process and dynamically selects the surrogate model most suitable for the current stage via the SISSO model selector, thereby enhancing the algorithm’s adaptability to different function characteristics. Experiments on CEC2022 test functions show that the proposed method achieves significantly lower mean errors and stable variances on most functions. Particularly on functions F10 and F11, the SISSO method reduces the mean error by more than an order of magnitude compared to methods like ElasticNet, Lars, Lasso, and Ridge, demonstrating excellent comprehensive performance.