The purpose of this assignment is to "implement a neural network (NN) from scratch, without using the pre-implemented NN modules in popular machine learning frameworks such as PyTorch and TensorFlow, and train it to perform regression or classification with a small dataset of their choice".

The implementation is provided in the small-nn package which we install in the experiments Jupyter notebook for experiments on the MNIST dataset.

To install small-nn, simply clone this repository and run:

cd aims-ml-nn
pip install .

At that point you should be able to reproduce the results in the experiments.ipynb Jupyter notebook.