The goal is to classify handwritten digits using the famous MNIST data set. Although the data set itself is done and dusted, I thought I would try out deep learning using the "H20 - Bleeding Edge package" to make better predictions.
Here is a link to the package : http://www.h2o.ai/h2o/why-h2o/
Applying machine learning techniques becomes incredible simple using this platform. And even the most basic framework yields an accuracy of 96% in Kaggle.