摘要： 为克服准遗传算法收敛速度慢、早熟收敛等缺点，提出一种方向自学习遗传算法，该算法在局部搜索中引入方向信息，利用函数的伪梯度来指导搜索方向。算法通过个体之间的竞争、合作与学习来不断更新最优个体，为增加种群的多样性提出一种消亡算子，避免早熟收敛，提高算法收敛速度。采用4个二维函数和多个无约束高维函数对算法进行测试，与3个新提出的算法进行比较，实验数据和理论分析表明，该算法在解的质量上和计算复杂度上都优于上述3个算法，充分证明该算法的有效性。

关键词: 遗传算法, 方向自学习, 函数优化, 进化计算

Abstract: In order to overcome the disadvantages of classical genetic algorithm of low convergence speed and avoid prematurity, this paper proposes an improved Genetic Algorithm of Directional Self-Learning(DSLGA). The directional information is introduced in local search process of the self-learning operator. And the search direction is guided by the false derivative of the function fitness. By the competition, cooperation and learning among the individuals, best solution is updated continuously. And a deletion operator is proposed to increase diversity. So the prematurity is avoided. In experiments, DSLGA is tested by four bi-dimension functions and three unconstrained benchmark problems, and the results are compared with CGA, MGA, FEP and OGA/Q. It shows that DSLGA performs much better than the above algorithms both in quality of solutions and in computational complexity. So the validity of the algorithm is obvious.

Key words: genetic algorithm, directional self-learning, function optimization, evolutionary computation

中图分类号: 

• TP301.6