Extremely efficient procedures for fitting regularization path with l0, l1, and truncated lasso penalty for linear regression and logistic regression models. This version is a completely new version compared with our previous version, which was mainly based on R. New core algorithms are developed and are now written in C++ and highly optimized.
