To clarify, my expectation is that this monster pass will let us do a fairly general kernel AND the SVD in one pass.
On Fri, Feb 26, 2010 at 12:08 AM, Jake Mannix <[email protected]> wrote: > stochastic decomposition work will certainly be the way > to go, as it allows you to do it in *one* monster pass over the data. > -- Ted Dunning, CTO DeepDyve
