摘要
提出了基于DNA计算和遗传算法的DNA遗传算法.给出了DNA遗传算法的结构,讨论了遗传操作算子.利用DNA遗传算法对FNN进行学习,比采用梯度型算法和遗传算法有更高的学习精度和更快的收敛速度.该算法有全局收敛性避免了采用梯度型学习算法训练FNN时固有的局部收敛问题,同样,该算法加速了FNN的训练,能够在线应用.
A DNA genetic algorithm is proposed based on DNA computing and genetic algorithms. The structure of tne DNA genetic algorithms is presented, and the operators of the proposed algorithm are discussed. Compared with the grads algorithm and the genetic algorithms,the DNA genetic algorithm owns more accurate precision and faster convergent speed when it is used to train FNN. The algorithm obtains the performance of global convergence which avoids the inherent convergence when the grids algorithm is adopted to train FNN. Also the proposed algorithm accelerates FNN＇ s training and can be used on-line.
出处
《小型微型计算机系统》
CSCD
北大核心
2006年第7期1270-1274,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60174047)资助.
