MNIST digit recognizer

Neural network based techniques to recognize digits from handwritten images in the MNIST dataset. There are two main implementations in this code.

1. network2.py has the code for a Neural network we create using just the numpy library.
2. mnist-pytorch.ipynb does the same task using pytorch

Setup to train neural network without Pytorch

Create a python environment:

$ python3 -m venv env

Activate the environment

$ source env/bin/activate

Install the requirements

$ pip install -r requirements.txt

Running Instructions

1. To train the model, and save it

$ python runner.py train

The model will be saved as mnist_model.py

2. To check the model output

$ python runner.py test

This will load the saved model. It will print the predicted number. The image for which the prediction was run will be saved in the project directory as infer.png.

References

1. http://neuralnetworksanddeeplearning.com/chap1.html
2. https://pytorch.org/tutorials/beginner/deep_learning_60min_blitz.html