摘要
Cuckoo search (CS) algorithm is a popular and efficient search technique for tackling numerical optimization problems. Nevertheless, CS algorithm is prone to premature convergence in solving complex multimode problems. Motivated by this observation, we propose a novel CS algorithm named multi-strategy adaptive CS (MACS). Specifically, in MACS, a parameter control strategy based on the Cauchy distribution and Lehmer mean is employed to dynamically update the step size. After that, three search strategies with different advantages are integrated to complement one another throughout the evolution process, which further strengthens the versatility and robustness. Besides, after every certain number of generations, a probability matching scheme is adopted to adaptively determine the best performing search strategy to produce more promising offspring. Extensive experiments are conducted on 42 benchmark problems from two different test suites. In comparison to seven advanced CS variants as well as other several popular algorithms, the experimental results indicate that MACS exhibits better overall performance.
学科与关键词
- Multi-strategy serial cuckoo search algorithm for global optimizationPeng, Hu*; Zeng, Zhaogan; Deng, Changshou; Wu, ZhijianKnowledge-Based Systems, 2021, 214: 106729.被引 49
- Parameter Control Based Cuckoo Search Algorithm for Numerical OptimizationCheng, Jiatang; Xiong, Yan*NEURAL PROCESSING LETTERS, 2022, 54(4): 3173-3200.被引 4
- A novel enhanced cuckoo search algorithm for global optimizationLuo, Wenguan; Yu, Xiaobing*JOURNAL OF INTELLIGENT & FUZZY SYSTEMS, 2022, 43(3): 2945-2962.被引 2
