论文部分内容阅读
在研究MADALINES人工神经网络中心差梯度学习算法(CDG)的基础上,提出了一种加速中心差梯度学习算法(SCDG),给出了该算法收敛性的证明,在计算机上模拟实现了SCDG算法且分析了实验结果,理论分析与实验均表明,SCDG算法较之传统的CDG算法,收敛速度提高了二个数量级以上.
Based on the study of the MADALINES artificial neural network center gradient difference learning algorithm (CDG), an accelerated center difference gradient learning algorithm (SCDG) is proposed and the convergence of the algorithm is proved. The SCDG algorithm is simulated on the computer The experimental results are also analyzed. The theoretical analysis and experimental results show that the convergence speed of SCDG algorithm is more than two orders of magnitude higher than that of the traditional CDG algorithm.