This repository contains the code for a programming assignment involving digit recognition using neural networks on a modified version of the MNIST dataset.
A description of the assignment can be found in the homework description.
The neural network can be run against a training and testing dataset in the data folder (train_image.csv, train_label.csv, test_image.csv, test_label.csv) through the following commands:
python NeuralNetwork3.pyThe player was tested against images from the MNIST dataset and those created by the course producers. A description of these test cases can be found in the homework description. The network performed to over 90% accuracy on the MNIST dataset, and over 55% accuracy on the hand-crafted dataset.