Welcome

 

I am currently participating in the Netflix Prize challenge. I am making improvement day by day, and hope one day I can catch up with the leaders.

Majority of techniques I use are similar to what others are using, SVD, KNN, RBM etc. Though I've found that some papers "hide" certain important information.  

Recommendation System is really one of the leading analytical techniques nowadays, and of course, will bring in new market opportunities.

The phrase, 'Dao Ci Yi You', means 'I have been here'. If you are reading this page, you are 'Dao Ci Yi You'!

Contact Me: BruceDeng@gamil.com

常用推荐系统算法 - 

1) SVD: SVD 算法在Netflix Prize中得到了广泛的应用,其中还衍生出了各种变种,如BellKor提出的Asymetric SVD, SVD++等等。 SVD主要特点就是运算迅速,预测简单,概念相对清楚理解,所有的这些都是其他算法无法比拟的。而且SVD经过简单改进后便可在后台运算,灵活的特点决定了将来商用的巨大价值。


2) RBM - RBM是多伦多大学的人工智能实验室提出来的,相比SVD,这个算法要难理解好多。但是这个算法有个优点,那就是独特的学习网络结构,以及不停的随机取样逼近,使得这个算法如果和其他算法结合,能够产生较大的进步。当然,算法的运算量也是惊人的,相比与SVD,RBM的运算时间是5到10倍。


3)KNN是另外一种古老的算法 - 但是在Netflix Prize 中有着新的运用,那就是KNN在其他算法后的后处理,以及相似度的动态学习产生的过程。

Sign in  |  Recent Site Activity  |  Terms  |  Report Abuse  |  Print page  |  Powered by Google Sites