This package implements the CLaR alorithm, a fast algorithm to handle sparse linear regression with heteroscedastic noise, see: Bertrand, Q., Massias, M., Gramfort, A., & Salmon, J. (2019). Concomitant Lasso with Repetitions (CLaR): beyond averaging multiple realizations of heteroscedastic noise. arXiv preprint arXiv:1902.02509.
It also implements a variation of the MRCE algorithm with a L21 penalization on the regression coefficient Beta, and which take in account the repetitions see Rothman, A. J., Levina, E., & Zhu, J. (2010). Sparse multivariate regression with covariance estimation. Journal of Computational and Graphical Statistics, 19(4), 947-962.
To be able to run the code you first need to run, in this folder:
`pip install -e .`
`python -c "import clar"`
should run smoothly, and you can execute the example.py file.
ArXiv link: https://export.arxiv.org/abs/1902.02509