Improved spherical search algorithm with memory-based dynamic population for optimization

Sicheng Liu, Sichen Tao, Kaiyu Wang, Zhenyu Lei, Shangce Gao

Abstract: The spherical search algorithm (SS) generates novel solutions by partitioning the population and utilizing a spherical search space. However, the fixed size of sub-populations leads to an accelerated convergence rate in SS, which often results in being trapped into local optima. This paper presents an advanced SS enhanced with a memory-based dynamic population scheduling system (SSDS). Building on the foundational SS framework, SSDS innovates with a dynamic population approach, utilizing a sub-population ratio record sequence to leverage historical data, and multiplexing historical population proportions reasonable improve the exploration behavior. As a result, SSDS dynamically balances exploration and exploitation throughout the search process. Preliminary results indicate that SSDS surpasses contemporary nine algorithms in IEEE congress on evolutionary computation (CEC) benchmark tests and exhibits promising application in 22 complex real-world problems. A closer analysis of the search performance and population diversity further highlights the effectiveness of the proposed SSDS.

External IDs:dblp:journals/tjs/LiuTWLG25