Convolutional Neural Network
Used dataset to train CNN without using any other data source or pre trained networks. Explained design choices in the ipython notebook itself during preprocessing, model building and training. Then the pretrained network is trained on MNIST dataset. Thourough analysis with the model checkpoints are saved. Comparison to a randomly initialized network in terms of convergence time, final accuracy and other possible training quality metrics is provided. Provided test accuracy on test set.