Research on improved convolutional wavelet neural network

Created on 2023-02-22T00:20:30-06:00

Return to the Index

This card pertains to a resource available on the internet.

This card can also be read via Gemini.

Wavelet transforms can replace standard backpropagation neurons and radial basis function based neurons to create "wavelet neural nets." These use a wavelet function as their activation function.

"Convolutional wavelet neural networks" take convolution neurons but replace fully connected layers with WNNs.

Wavelet based neural networks train faster (1/th speed of standard backpropagation) and can be slightly more accurate (~2%) than baseline CNNs on the MNIST dataset.

WNN is designed as follows: Firstly, structure of BPNN is adopted as the basic structure of WNN; Secondly, the form of activation function in hidden layers of RBFNN is adopted; Thirdly, the wavelet transform function is adopted as the activation function.