Express Wavenet: A lower parameter optical neural network with random shift wavelet pattern

Express Wavenet is an improved optical diffractive neural network. At each layer, it uses wavelet-like pattern to modulate the phase of waves. For input image with n2 pixels, express wavenet reduce parameter number from O(n2) O(n). Only need one percent parameters, and accuracy still very high. In MNIST dataset, only needs 1229 parameters get 92%, while standard network 125440 parameters. The random shift wavelets show characteristics more vividly. Especially vanishing gradient phenomenon in training process. We present a modified expressway structure for this problem. Experiments verified effect wavelet structure. Our work shows would use much fewer than other networks. source codes are available at https://github.com/closest-git/ONNet.