On Thu, Mar 5, 2009 at 4:24 AM, Sean Owen <[email protected]> wrote: > This would be a fantastic project, implementing a Recommender based on > this approach . I tried implementing an SVD technique a couple years > ago and it was waaay too slow on one machine. Revisiting with Hadoop > sounds great.
SVD (at least how it is used for CF---matrix approximation) is really just least squares, which is not hard (quite simple if you have a Non-linear Conjugate Gradients solver) and easy to regularize to get even better answers. I'd be happy to discuss further if anyone is interested (did a large portion of my PhD work in this area). Cheers, Jason -- Jason Rennie Research Scientist, ITA Software http://www.itasoftware.com/
