IEEE Access (Jan 2021)
Adaptation of Population Size in Sine Cosine Algorithm
Abstract
Sine Cosine Algorithm (SCA) is a newly proposed competent population-based metaheuristic, which has gained a multi-disciplinary interest in solving optimization problems. Like other metaheuristics, the performance of SCA is sensitive to the settings of its parameters and one such parameter is Population Size (PS). There is no one population size that fits all; problem uniqueness requires a matching strategy of parameter selection and adaptation. However, in standard SCA and its variants, PS is treated as a user-controlled parameter and no study has explored the effect of PS adaptation on SCA’s performance. To fill this research gap, this study investigates and compares the impact of promising strategies for setting and controlling population size, from other metaheuristics, on the performance of the standard SCA and five of its variants in terms of fitness, run-time, and convergence characteristics. Finding the best PS setting for a metaheuristic is a challenging problem since it depends on both the nature of the algorithm used and the problem being solved. Leading PS adaptation techniques considered in this study are linear staircase reduction, iterative halving, reinitialization and incrementation, pulse wave, population diversity, and three parent crossover strategies. A classic set of 23 well-known benchmark functions has been utilized for a fixed number of evaluations to assess the impact of each PS adaptation strategy on the performance of SCA and its variants. Also, non-parametric Wilcoxon’s rank sum test is performed to provide a comprehensive view of various PS adaptation strategies’ performance with respect to each other in terms of fitness and run-time. Simulation results reveal that proper selection of PS adaptation strategy can further enhance the exploration and exploitation capabilities of SCA and its variants.
Keywords
