This is the code for our project at Princeton (co-authors: Yuqi Nie, Hui Yuan). The paper title is: "Neural Network is heterogeneous: Phase Matters More". To run the program you need to install tensorflow. Then you may activate your tensorflow environment and import MNIST as we have done here.
Mnist-complexNN-final.py: This is the code for complex NN. It has a single hidden layer with 100 nodes, and the back-propagtion is redesigned according to the appendix.
Mnist_realNN_multilayer-final.py: This is the code for real NN. It has two hidden layer with 512 and 100 nodes separately.