I got this "axis of interest" concept from a presentation by one of the Netflix team runner-ups, I don't know which one. He did not give a name for it. Is there a standard term? I hate just making up new words.
Also, there are clusters of items at both ends, but there are also items along the axis which form a spectrum from one end to the other. So you get a line of items with blobs at the ends. On Fri, Aug 26, 2011 at 2:21 AM, Sean Owen <[email protected]> wrote: > That's correct. Well you just have to recompose the user row you are > interested in. It will no longer be sparse, at all. Those new values are > your estimated ratings. > > On Fri, Aug 26, 2011 at 12:07 AM, Jeff Hansen <[email protected]> wrote: > > > > I also think I may have missed a big step of the puzzle. For some reason > I > > thought that just by loosening the rank, you could recompose the Matrix A > > from the truncated SVD values/vectors and use the recomposed values > > themselves as the recommendation. I thought one of the ideas was that > the > > recomposed matrix had less "noise" and could be a better representation > of > > the underlying nature of the matrix than the original matrix itself. But > > that may have just been wishful thinking... > > > > > -- Lance Norskog [email protected]
