Machine Learning project for the Digit Recognizer competition in Kaggle.
This project doesn't use the full MNIST dataset, only the part provided by the Kaggle competition.
The algorithm current accuracy for the competition is 0.99414.
Requirements.txt contains the packages to install with pip. I have only tested it on Python 3, no guarantees it will work for Python 2.
The project can be executed through 'main.py', which imports 'utils.py', or as a single script with 'script.py'. This last script can be found and executed in Kaggle.