Genetic Programming-Based Evolutionary Strategy Generation for Complex Optimization

Jeng-Shyang Pan, Wenda Li, Shu-Chuan Chu, Zhi-Gang Du, Hongmei Yang, lingping kong

Keywords: Metaheuristic optimization, Genetic Programming, Adaptive update strategy

Abstract: Designing effective update strategies is vital to the success of metaheuristic algorithms. Traditional methods often depend on manually designed rules and empirical tuning, which restrict their adaptability and scalability across different optimization problems. To overcome this limitation, a novel framework named GP-MAs is proposed, which integrates Genetic Programming (GP) into metaheuristics to automatically evolve and optimize their update equations. In this framework, GP dynamically constructs learning rules within the search process, allowing the algorithm to adapt to varying problem landscapes. The framework is implemented on the Growth Optimizer (GO), forming a hybrid variant called GPbasedGO. Experimental evaluations on the CEC2022 benchmark suite demonstrate that GPbasedGO achieves superior convergence speed, robustness, and generalization ability compared with several state-of-the-art algorithms. The proposed GP-MAs framework offers a flexible and automated paradigm for metaheuristic design, enabling the generation of adaptive update strategies suitable for complex optimization tasks and real-world applications.

Submission Number: 78