Introduction to the genlasso package

We present a short tutorial and introduction to using the R package genlasso, which is used for computing the solution path of the generalized lasso problem discussed in Tibshirani and Taylor (2011). Use cases of the generalized lasso include the fused lasso over an arbitrary graph, and trend fitting of any given polynomial order. Our implementation includes a function to solve the generalized lasso is its most general form, as well as special functions to cover the fused lasso and trend filtering subproblems. The general implementation maintains and updates a matrix factorization to successively solve related linear systems; the specialized implementations forego this update routine and exploit subproblem structure, which improves both stability and speed of the computation. Standard S3 methods such as plot, predict, and coef are also included to assist in studying the output solution path objects.