摘要
针对目前局部回归神经网络动态ＢＰ算法的误差导数计算复杂、收敛速度慢的缺陷，提出了一种新的快速算法．该算法是将信号流图引入动态ＢＰ算法，较好地解决了求解误差导数的复杂性，同时采用ＢＦＧＳ算法加快了网络的收敛速度．仿真结果表明了本算法的有效性．
A new fast learning algorithm for recurrent neural networks is proposed. By introducing the signal flow graphs technique, it overcomes the disadvantage of complexity of the gradient of the error function. And for more fast convergence, the BFGS method is used. Simulation results show that the proposed algorithm converges faster than the traditional algorithm.
出处
《北京科技大学学报》
EI
CAS
CSCD
北大核心
2000年第1期89-92,共4页
Journal of University of Science and Technology Beijing
基金
国家"八五"攻关资助!No.85-311-02-11-04
