myCNN.py:
the classes and functions defined for CNN. This is a naive CNN, with no tricks(Dropout, Momentum, Mini-batch, GPU etc.) in it.
CNN_for_MNIST.py:
an example applied on the MNIST dataset. This network can overfit a small subset of MNIST. Considering the time, I didn't train this model on the whole dataset (and I won't, either). If you are interested in it, you can optimize it and inform me your result.
You can download dataset from:
https://www.kaggle.com/c/digit-recognizer/data