Deep Learning with MNIST Dataset

This is a project that gains high-level accurate predictions of numerical results using the MNIST dataset. It runs the model through the following processes:

1. A simple linear model
2. A fully connected, single layer neural network
3. A convolutional neural network (CNN)
4. A CNN with Batch Normalization

Dropout and Data Augmentation techniques are used to control for overfitting. Basic ensembling is used to achieve the best possible final results.

Installation

1. Run the requirements file
2. Open MNIST_Jakes.ipynb, and follow the process

Requirements

A GPU is needed to run this notebook - an AWS P2 instance works well.