R/Backsolve.R

In statsmaths/genlasso: Path Algorithm for Generalized Lasso Problems

# Convenience function to find the minimum 2-norm
# solution of
#               min_x  ||b - A^T x||_2
# when A is upper triangular, but has the first q
# columns equal to zero.

Backsolve <- function(A,b,q) {
  m = nrow(A)
  n = ncol(A)
  z = backsolve(A[Seq(1,n-q),Seq(q+1,n),drop=FALSE],
    b[Seq(q+1,n)],transpose=TRUE)
  return(c(z,rep(0,m-n+q)))
}