A novel metaheuristic algorithm: advanced social memory optimization
Shijie Fan, Ruichen Wang, Kang Su
Abstract
Abstract The field of optimization problems has garnered significant attention due to its importance across various applications, particularly driven by the demand for efficient solutions to complex engineering challenges. Numerous metaheuristic algorithms inspired by animal behaviour or swarm intelligence have been proposed; however, these algorithms often pursue a multitude of strategies, resulting in excessive parameters that complicate tuning and hinder convergence and balance. Additionally, algorithms based on human behaviour remain scarce. To address these limitations, extensive research has been conducted on the interplay between social memory and individual memory, leading to the introduction of a novel human-behaviour-inspired metaheuristic algorithm, named Advanced Social Memory Optimization (ASMO). This algorithm seeks to address the complexities of parameter management, convergence, and balance more effectively with a streamlined set of strategies. Furthermore, a mathematical model based on the mechanisms of social memory formation and individual memory updating underpins the algorithm. Rigorous performance evaluations, utilizing the Wilcoxon Rank-Sum Test and the Friedman Test across multiple benchmark suites (CEC2017, CEC2019, and CEC2022), demonstrate that ASMO, with only two algorithmic strategies, outperforms or matches established algorithms on more than half of the test functions. These findings suggest promising new avenues for research in the field of optimization and, given the succinctness of ASMO’s strategies, underscore its potential as a powerful tool for enhancing and developing solutions to complex engineering design problems. The code for the ASMO algorithm is available in appendix D.