Southeast University / Bachelor
Implementation of frequency domain convolutional neural network
On the basis of the existing convolutional neural network, fractional-level Fourier transform is introduced to process the original signal. Multiplication operation in the transform domain is used to replace the original convolution operation, and the corresponding pooling layer is designed. A three-layer convolutional neural network is built based on the Tensorflow platform and its effect is verified by the MNIST data set. Responsible for technical research and arrangement in the project. The project was rated good.