Feedforward neural network's denoising with wavelet basis
Jianwei Li, Zhong Chengge, Dong Huachun, Quan Taifan · 2002
Methods based on wavelet transform theory for decreasing sampling noise in feedforward neural networks are proposed in this paper. Wavelet bases are employed in the network to constrain the network's ability in learning samples which are corrupted by noise. The selection of the wavelet bases which correlate with the map to be approximated is mainly discussed.