This repository conatains various approach used to classify digits in the Digit Classifier Competition of Kaggle.

The approach consist of:

1. HOG Features with Linear SVC which gave an accuracy of 97.5%.
2. HOG features with simple 2 Layer Neurak Network giving an accuracy of 98.2%.
3. Convolution Neural Networks with Max Pooling and Dropout Giving an accuracy of 99.3%.