yxwang99/alphanorm: alpha-norm regularization

An implementation of alpha-norm regulariztaion linear model in R. The alpha-norm penalty has the property of jumping to a sparse solution. This flexible nonconvex regularization problem is solved via cyclic coordinate descent and a proximal operator. It is less aggresive in shrinking coefficients than the l_0 penalty , sparser and less biased than l_1 norm(lasso), which is extremely useful in high-dimensional case and when predictors are highly correlated. Our package also offers the choice of lasso (q=1), it can be useful when the model is not extremely sparse.

install.packages("remotes")
remotes::install_github("yxwang99/alphanorm")