Netflix Prize Forum's Summary(4/25/07)

1. Netflix Recommender Framework
Icefox，免費提供c++ source code(我有試過,可是不能RUN,我ㄉ電腦有些它需要軟體好像沒有),而且最近一版ㄉCODE 裡,有提供SVD algorithm (singular value decomposition (SVD))

http://www.netflixprize.com/community/viewtopic.php?id=352
http://sifter.org/~simon/journal/20061211.html (這個是寫SVD 那個人ㄉ網頁)

2. Pyflix - The pythonic Netflix Prize
Deus ex silicon，免費提供python source code

http://www.netflixprize.com/community/viewtopic.php?id=432

3. Useful public RMSEs
一些基本ㄉ猜測ㄉRMSE(像是全猜1或2)
下面這個怪公式ㄉRMSE 可以低於1
For the following equation:
R = -2.58797372 + 0.95682133*M + 0.7606549*U

Where:
R = predicted rating
M = movie average
U = user average

http://www.netflixprize.com/community/viewtopic.php?id=435

4.Not sure how to calculate RMSE
如何計算RMSE

http://www.netflixprize.com/community/viewtopic.php?id=625

5.A better spreadsheet than Excel ?
這個人用EXCEL
http://www.netflixprize.com/community/viewtopic.php?id=483

6.Anyone help with Collaborative
算兩個userㄉ相關

http://www.netflixprize.com/community/viewtopic.php?pid=4924#p4924

7. Ramblings on the lower bound on the RMSE
http://www.netflixprize.com/community/viewtopic.php?id=78

RMSE-------perfect%----random%
0.2887-----100%--------0%
0.8563-----48%---------52%
0.9514-----34%---------66%
0.98--------30%---------70%
1.0540-----18%---------82%
1.1547-----0%---------100%

8.Why RMSE=1.79 for random ratings?
亂猜也有RMSE=1.79

http://www.netflixprize.com/community/viewtopic.php?id=224