This project is about performing feature selection in neural networks. At the moment, we support fully connected feed-forward neural networks. LassoNet is based on the work presented in this paper (bibtex here for citation). Here is a link to the promo video:
We have designed the code to follow scikit-learn's standards to the extent possible (e.g. linear_model.Lasso).
To install it,
pip install lassonet
Our plan is to add more functionality that help users understand the important features in neural networks.
LassoNet's website is https://lassonet.ml. It contains many useful references including the paper, live talks and additional documentation.