On Sun, Dec 13, 2009 at 3:24 AM, Jake Mannix <jake.mannix@gmail.com> wrote:
> You do the co-occurrence matrix (for item-by-item, right?) on Hadoop too,
> and that part is really fast, but computing the recommendations is very
> slow? By what orders of magnitude, for the whole set?
>
> What are the scales you are testing with, in terms of total number of users,
> items, and ratings?
Yes, for about 10M ratings (tens of thousands of users and items) the
co-occurrence matrix counts take a couple minutes, and then recs are
on track to take a day or two.
|