关键词: 高维多模态问题, 自适应差分进化, 高斯分布估计算法, 小生境, 文化算法

Abstract:

Aiming at the defects of slow rate of convergence and easily falling into local optimum in the traditional evolution algorithm, a selfadaptive Cultural Algorithm (CA) based on Differential Evolution (DE) and niche elite Gaussian Estimation of Distribution Algorithm is proposed to resolve highdimensional multimodal optimization problems. The selfadaptive differential evolution algorithm is used to optimize the population space and the niche elite population is recognized by dynamic recognition algorithm. In the belief space, the niche elite population is optimized by Gaussian Estimation of Distribution Algorithm. The optimized result and the size and characteristics of the niche are stored into the evolution knowledge base. Then, the population in the population space is guided and inspired by the evolution knowledge base. It guarantees population diversity and avoids the duplication of local search. Finally, this algorithm is tested on 4 multimodal benchmark functions, and the experimental result shows the algorithm has advantages in convergence velocity, solution precision, stabilization and global search capability.

Key words: highdimensional multimodal;adaptive differential evolution;Gaussian estimation of distribution algorithm;niche elite;cultural algorithm