Are there any good papers or books dealing with the use of coordinate descent for L1 (lasso) and/or elastic net regularization for linear regression problems?
I earlier suggested the recent paper by Friedman and coll., Regularization Paths for Generalized Linear Models via Coordinate Descent, published in the Journal of Statistical Software (2010). Here are some other references that might be useful:
- Pathwise coordinate optimization, by Friedman and coll.
- Fast Regularization Paths via Coordinate Descent, by Hastie (UseR! 2009)
- Coordinate descent algorithms for lasso penalized regression, by Wu and Lange (Ann. Appl. Stat. 2(1): 224-244, 2008; also on available on arXiv.org)
- Coordinate Descent for Sparse Solutions of Underdetermined Linear Systems of Equations, by Yagle (a bit too complex for me)
I've just come across this lecture by Hastie and thought that others might find it interesting.