用前馈神经网络进行带噪声信号的去噪声建模

本文提出了一种用前馈神经网络对带噪声样本的去噪声建模的实验方法，能获得合适的网络模型，并具有较好的去噪声能力。实验对比ＢＰ网络和ＲＢＦ网络去噪声能力上的差别，结果表明，ＲＢＦ网络去噪声能力优于ＢＰ网络。这一结论已被用于为半导体生产工艺控制参数优化的去噪声建模中。

Feed-forward Neural Network Modeling for Noise Rejection

An empirical method was proposed to model the nonlinear system using the noise-contaminated samples, which can be used to determine a proper network model and achieve better noise elimination performance. The comparison of the noise-filtering capacity of the back-propagation network (BPN) and the radial basis neural function network (RBFN) is also presented and the experimental results show that the RBFN is superior to the BPN. The method has been successfully applied in modeling and optimization of the control parameters of IC manufacturing processes.