The lasso can be computed with the LARS or Coordinate Descent algorithm.
What is their computational complexity and when one is quicker than the other?
The lasso can be computed with the LARS or Coordinate Descent algorithm.
What is their computational complexity and when one is quicker than the other?
I realize it's quite late to give an answer but maybe someone will find it useful.
Here is a nice talk by Trevor Hastie about the coordinate descent. He compares (among others) his two R packages: glmnet
(using coordinate descent) and lars
(using LARS).
It's shown that coordinate descent is faster in each setting: $p>N, p<N$ and sparse or dense data. There are some examples featuring both simulated and real data.