Fast iterative regularization approach for non strongly convex regularizers, e.g. sparsity (L1) or low rank (nuclear norm).
Please visit https://LCSL.github.io/iterreg/ for the latest version of the documentation.
From a console or terminal clone the repository and install iterreg:
git clone https://github.com/LCSL/iterreg.git cd iterreg/ pip install -e .
To build the documentation you will need to run:
cd doc/ pip install doc_requirements.txt make html
In the example section of the documentation, you will find scripts to reproduce the figures of the paper.
All dependencies are in the ./requirements.txt file.
They are installed automatically when pip install -e . is run.
If you use this code, please cite:
@inproceedings{iterreg,
title = {Iterative regularization with convex regularizers},
author = {Molinari, Cesare and Massias, Mathurin and Rosasco, Lorenzo and Villa, Silvia},
booktitle = AISTATS,
year = 2021,
}
ArXiv links: