论文部分内容阅读
该文提出一种适用于单层神经网络（ＳＮＮ）训练的新颖的广义误差函数，给出了ＳＮＮ新的快速学习算法（ＦＬＡ）．进一步提出了一种广义系统辨识模型，对ＦＬＡ的收敛性进行了理论分析．实验表明：文中给出的新ＦＬＡ比Ｋａｒａｙｉａｎｎｉｓ的ＬＦＡ具有更快的收敛速度
In this paper, a novel generalized error function suitable for single-level neural network (SNN) training is proposed, and a new fast learning algorithm (FLA) for SNN is given. Furthermore, a generalized system identification model is proposed and the convergence of FLA is analyzed theoretically. Experiments show that the new FLA given in this paper has a faster convergence rate than Karayiannis’ s LFA