py-CNN
The main file in this project is nonMNIST-CNN.py, which is trained on the majority of the images in the 'data/' folder (after which the training data is saved in a .csv format) and then tested on the remaining data. In practice, this neural network was able to correctly identify non-trivial examples of written symbols with over 85% accuracy.
The remainder of the files are retained for the sake of documenting how and why this project was performed, as well as providing some visual feedback on the algorithm's accurary in graph form.