Re: Linear CG solver

2014-06-28 Thread Tom Vacek
> of them running on datasets > > I am hoping we can re-use the solver for SVM variants... > > Could you please point to some implementation references for Nystrom > approximations or kernel mappings ? > > Thanks. > Deb > > > On Sat, Jun 28, 2014 at 11:26 AM, To

Re: Linear CG solver

2014-06-28 Thread Tom Vacek
What flavor of SVM are you trying to support? LSSVM doesn't need a bound constraint, but most other formulations do. There have been ideas for bound-constrained CG, though bounded LBFGS is more common. I think code for Nystrom approximations or kernel mappings would be more useful. On Fri, Jun

Re: MLlib - logistic regression with GD vs LBFGS, sparse vs dense benchmark result

2014-04-25 Thread Tom Vacek
I don't know about Spark's implementation, but with LBFGS, there is a line search step. Since computing the line search takes roughly the same work as one iteration, an efficient implementation will take a full step and simultaneously compute the gradient for the next step and check if the update