基于U－D分解推广卡尔曼滤波的神经网络学习算法

本文针对前馈神经网络ＢＰ算法所存在的收敛速度慢区常遇局部极小值等缺陷，提出一种基于Ｕ－Ｄ分解的渐消记忆推广卡尔曼滤波学习新算法．与ＢＰ和ＥＫＦ学习算法相比，新算法不仅大大加快了学习收敛速度、数值稳定性好，而且需较少的学习次数和隐节点数即可达到更好的学习效果，对初始权值，初始方差阵等参数的选取不敏感，便于工程应用．非线性系统建模与辨识的仿真计算表明，该算法是提高网络学习速度、改善学习效果的一种非常有效的方法．

A New Fast Learning Algorithm for Feedforward Neural Networks Using U-DFactorization-Based Extended Kalman Filter

A New fast learning algorithm for training multilayer feedforward neural networks by using variable time--varying forgetting factor technique and U-D factorization based fading memory extended Kalman filter is proposed in this paper. In comparison with BP and ex tended Kalman filter (EKF ) based learning algorithm, the new algorithm can not only obviously improve the convergency rate,numerical stability. but also provide much more accurate learning results in fewer iterations with fewer hidden nodes. In addition, it is less affected by the choice of initial weights and initial covariance matrix as well as other setup parameters. The results of simulated computations of nonlinear dynamic system modelling and identification applications show that the new algorithm proposed here is an effective and efficient learning algorithm for feedforward neural networks.