Abstract: To solve the shortcomings of sparrow search algorithm (SSA), such as falling into local extremum easily influenced by initial solution and slow convergence in late iteration, a multi-chaotic sparrow search algorithm  based on learning mechanism (MMCSSA) is proposed. Firstly, the centroid opposition-based learning strategy (COBL) is introduced to generate elite population to enhance the exploration of multi-source high-quality search areas, and then the local extreme escape ability and convergence performance of the algorithm are improved. Secondly, a scaled golden sine algorithm is proposed and embedded in SSA to improve the guidance search mode and enhance the global search ability of the algorithm. Thirdly, a multi-chaos mapping strategy based on learning mechanism is proposed, which utilizes the characteristics of multi-chaos and multi-disturbance, and enforces different disturb-ance features on the algorithm by dynamically calling different chaotic maps. When chaotic perturbation fails, Gaussian mutation strategy is introduced to deeply develop the current solution. The two strategies cooperate and promote each other, greatly enhancing the ability of the algorithm to escape from local optimal. Finally, the proposed algorithm is applied to 12 benchmark functions with different characteristics, and the results show that compared with other algorithms, MMCSSA has better performance in convergence accuracy, optimization speed and robustness.

Key words: sparrow search algorithm (SSA), golden sine algorithm, Gaussian mutation, multiple chaotic learning mechanisms, centroid opposition-based learning strategy (COBL)

摘要： 针对麻雀搜索算法（SSA)易受初始解的影响陷入局部极值、迭代后期收敛速度慢等缺陷，提出了一种融合学习机制的多混沌麻雀搜索算法（MMCSSA）。首先，引入重心反向学习策略（COBL）生成精英种群增强对多源优质搜索区域的勘探能力，提升算法的局部极值逃逸能力和收敛性能。其次，提出一种动态调整的黄金正弦领导策略并嵌入SSA中以改善发现者的搜索方式，增强算法的全局搜索能力。然后，提出一种基于学习机制的多混沌映射策略，该机制利用多混沌多扰动模式的特性，通过动态调用不同混沌映射赋予算法不同类别的扰动特征。混沌扰动失败时，引入高斯变异策略对当前解进行深度开发，两种策略协同作用，相互促进，极大增强了算法逃逸局部最优的能力。最后，将所提算法应用于12个不同特征的基准函数进行实验，结果表明与其他算法相比，MMCSSA在收敛精度、寻优速度和鲁棒性方面有更好的表现。

关键词: 麻雀搜索算法（SSA）, 黄金正弦算法, 高斯变异, 多混沌学习机制, 重心反向学习策略（COBL）