一种基于最优近邻交叉的遗传算法

A new genetic algorithm based on the best neighborhood crossover

在自然界中,个体总是企图向其周围优秀的个体学习,同时,这种非监督的学习方式不可避免的受环境的影响,因此,变异总是存在的。受这种现象的启发,为避免GA的早熟和收敛到局部极小,提出了一种新的优化算法,种群大小为N×N,均匀分布在一个正方形内,除边界外,每个个体都有8个邻居。在一个3×3邻域内,保持最优个体不变,其它个体同最优个体交叉,子代经过概率1变异后,取代原来非最优的父代。理论及函数寻优和模板匹配证明,该算法具有很强的寻优能力和很快的收敛速度。

In the nature, an individual is always impacted by her neighbor, and the individual always wants to learn from her best neighbor, and the learning is not supervised, so mutation is always influenced. In order to avoid premature convergence and occurrence of minimal deceptive problems, we introduced a new op- timal algorithm, the seeds distribute uniform in a square and its number is N × N. Except the seeds on the boundary, all the other seeds have eight neighbors. In the areas of 3 × 3, the optimal individual exists, while the others intercross with the optimal. After mutating in probability one, the child replaced the inferior parent. Theoretically, module matching of image and function optimize proved that this algorithm has more ability of looking for the optimal and faster of convergence.