Natural coordinate descent algorithm for L1-penalised regression in generalised linear models

Tom Michoel

The problem of finding the maximum likelihood estimates for the regression coefficients in generalised linear models with an L1 sparsity penalty is shown to be equivalent to minimi…