see the draft please! ----- Original Message ----- From: "DB Tsai" <dbt...@dbtsai.com> To: dev@mahout.apache.org Cc: "Ted Dunning" <ted.dunn...@gmail.com> Sent: Sunday, July 21, 2013 2:38:01 PM Subject: Re: [jira] [Comment Edited] (MAHOUT-1273) Single Pass Algorithm for Penalized Linear Regression on MapReduce
Coordinate decent is essentially a iterative algorithm, how can you do it in single pass of data with L2 regularization? On Jul 21, 2013 2:09 PM, "Michael Kun Yang" <kuny...@stanford.edu> wrote: > I will update the document to detail the algorithm. > > > On Sun, Jul 21, 2013 at 1:50 PM, Ted Dunning <ted.dunn...@gmail.com> > wrote: > > > On Sun, Jul 21, 2013 at 1:41 PM, Kun Yang <kuny...@stanford.edu> wrote: > > > > > The algorithm is not solving the normal equation as in the ordinary > > linear > > > regression. I did not detail the algorithm to solve the penalized > > > optimization in the paper. To solve the penalized version, I will use > the > > > coordinate descent which is well documented in other paper (see > > Freedman's > > > paper, for 1000 variables, it takes ~1min to do cross validation in > the R > > > package) and is very efficient. > > > > > > As I discussed in the conclusion section, to solve the problem with > large > > > number of predictors, it is still a challenge even though in the single > > > machine or MPI version, but the proposed algorithm can handle the > number > > of > > > variable at the order of 5000 and it covers lots of applications. > > > > > > > Should the document be updated to describe what you intend to do? > > >
