摘要
针对BP神经网络学习速度慢、容易陷入局部极小的缺点,提出了一种基于改进免疫遗传算法的多层前向神经网络,将该算法用于多层前向神经网络的权值优化,扩大了神经网络的权值搜索空间,提高了网络系统的学习效率和精度。将该神经网络用于上证指数的趋势预测,仿真结果表明:该神经网络比BP神经网络具有更好的全局收敛性、更高的学习效率和预测精度。
A multilayer feed-forward neural networks based on an improved immune-genetic algorithm (IIGA) is presented in this paper, so as to overcome the shortcomings that back propagation neural networks (BPNN) is easily converged at a local minimum and the training process is slow. The improved immune-genetic algorithm is introduced to optimize the weight of multilayer feed-forward neural networks. It enlarges the weight's search space and improves the training efficiency and precision of neural networks. The neural networks is adopted to forecast Shanghai stock indexes. The result of emulation indicates that the neural networks has better global convergence and higher training efficiency and forecasting precision.
出处
《华东理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2006年第11期1342-1345,共4页
Journal of East China University of Science and Technology
基金
国家重点基础研究发展规划项目(2002CB312200)
高等学校博士点专项科研基金项目(20040251010)
关键词
神经网络
免疫遗传算法
信息熵
亲和度
neural networks
immune-genetic algorithm
information entropy
affinity