A neural network with three hidden layers, and a self-implemented back propagation, to identify handwritten digits from the MNIST database with ~94% accuracy, and a best of 95% accuracy.
- Download the binary files of the test data from the MNIST database
- Run
python mnistToText.py
on both the training set and validation set to convert both files into text files. - Run
python app.py
to train the neural network on the training set, and then test its accuracy on the validation set.