celer 0.6

Fast algorithm with dual extrapolation for sparse problems

celer

Fast algorithm to solve Lasso-like problems with dual extrapolation. Currently, the package handles the following problems:

• Lasso

• weighted Lasso

• Sparse Logistic regression

• Group Lasso

• Multitask Lasso.

The estimators follow the scikit-learn API, come with automated parallel cross-validation, and support both sparse and dense data, with optionnaly feature centering, normalization, and unpenalized intercept fitting. The solvers used allow for solving large scale problems with millions of features, up to 100 times faster than scikit-learn.

Documentation

Please visit https://mathurinm.github.io/celer/ for the latest version of the documentation.

Install the released version

Assuming you have a working Python environment, e.g., with Anaconda you can install celer with pip.

From a console or terminal install celer with pip:

pip install -U celer

Install and work with the development version

From a console or terminal clone the repository and install Celer:

git clone https://github.com/mathurinm/celer.git
cd celer/
pip install -e .

To build the documentation you will need to run:

pip install -U sphinx_gallery sphinx_bootstrap_theme
cd doc/
make html

Demos & Examples

In the example section of the documentation, you will find numerous examples on real life datasets, timing comparison with other estimators, easy and fast ways to perform cross validation, etc.

Dependencies

All dependencies are in the ./requirements.txt file. They are installed automatically when pip install -e . is run.

Cite

If you use this code, please cite:

@InProceedings{pmlr-v80-massias18a,
  title =    {Celer: a Fast Solver for the Lasso with Dual Extrapolation},
  author =   {Massias, Mathurin and Gramfort, Alexandre and Salmon, Joseph},
  booktitle =        {Proceedings of the 35th International Conference on Machine Learning},
  pages =    {3321--3330},
  year =     {2018},
  volume =   {80},
}

@article{massias2019dual,
title={Dual Extrapolation for Sparse Generalized Linear Models},
author={Massias, Mathurin and Vaiter, Samuel and Gramfort, Alexandre and Salmon, Joseph},
journal={arXiv preprint arXiv:1907.05830},
year={2019}
}

ArXiv links:

• https://arxiv.org/abs/1802.07481

• https://arxiv.org/abs/1907.05830