tag:blogger.com,1999:blog-6569681.post56359219297151222..comments2024-03-29T05:14:10.903-07:00Comments on Geeking with Greg: Netflix Prize at KDD 2008Greg Lindenhttp://www.blogger.com/profile/09216403000599463072noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-6569681.post-23164921034906418932008-10-14T12:15:00.000-07:002008-10-14T12:15:00.000-07:00The surprisal value of a rating is to some degree ...The surprisal value of a rating is to some degree based on the previous ratings of the rating user, and to some degree the ratings of other users. A rating with no comparison or trending index is an unattached scalar. Coming up with the correct *relative* metric for a rating is definitely the best way to extract information from noise. I liked the idea that ratings might have the baseline index of "the users mood" on a particular day. Very interesting.Dr. Mark J. Boydhttps://www.blogger.com/profile/02407924268767054414noreply@blogger.com