Regularized neural network for identification of handwritten single digits

Notes

This was originally a repo for a uni project, but the master branch might contain changes made after such venture, including, but not limited to, translation and contribution from more or less people.

Dependencies

• python3, version >= 3.10

  • pip

  • venv

Instructions

Clone the repo and do:

python3 -m venv .env
source ./.env/bin/activate
pip install -r requirements.txt
python -m ipykernel install --name=.env
jupytext main.py --to notebook
jupyter lab main.ipynb

Then execute the notebook cells.

If any changes that should be kept are made, save the file and do:

jupytext main.ipynb --to py:percent

Otherwise, simply open the notebook again in another ocasion.

Credits

The function theta_meaning is heavily inspired in https://stackoverflow.com/questions/41793931/plotting-images-side-by-side-using-matplotlib/66961099#66961099.