论文部分内容阅读
本文提出了一种复数赫布类型的学习算法，用来训练一个输入、输出及权值均为复数的复神经网络．本算法的推导是基于ＴＬＳ准则，而不是常用的ＬＳ或ＬＭＳ准则．算法的收敛性及收敛结果通过将其应用到自适应复数ⅡＲ滤波器中来加以说明．
In this paper, we propose a complex Herb-type learning algorithm for training a complex neural network with input, output and weights as complex numbers. The derivation of this algorithm is based on the TLS criterion, not the commonly used LS or LMS criterion. The convergence and convergence of the algorithm are illustrated by applying it to an adaptive complex IIR filter.