Research on improving the alpha evolution optimization algorithm using hybrid strategies

Abstract To address issues such as premature convergence and entrapment in local optima during the optimization process of the Alpha Evolution (AE) algorithm, this study proposes an improved hybrid Alpha Evolution algorithm (HAE). The innovations of HAE are realized in three aspects: First, it adopts a two-dimensional cellular topology (Von Neumann-L5 neighborhood) to reconstruct individual interaction patterns, thereby maintaining population diversity via restricted information propagation; Second, a dynamic threshold-based hybrid search strategy is introduced to adaptively adjust the balance between global exploration and local exploitation; Third, a dynamic opposition-based learning mechanism is integrated, which generates high-quality opposite solutions using real-time search information to enhance the algorithm’s ability to escape local optima. Comparative experiments on the CEC2017 standard test functions and four typical engineering design problems show that the HAE algorithm outperforms the original AE algorithm significantly in terms of solution accuracy and convergence performance.