LFD Book Forum (http://book.caltech.edu/bookforum/index.php)
-   Homework 1 (http://book.caltech.edu/bookforum/forumdisplay.php?f=130)
-   -   question on netflix example (http://book.caltech.edu/bookforum/showthread.php?t=4154)

 arun.n 04-05-2013 06:42 PM

question on netflix example

Hello everyone

The way I understood the Netflix example is that, given a set of users and movies along with the ratings that the users gave to those movies, the learning algorithm learns the vectors for the profiles of users and movies. So the hypothesis is the set of vectors for the users and movies.

I can see how this can be used to predict the rating that a user would give to a movie he/she didn't watch provided the user and the movie were in the original dataset.

Now my question is how can this be used for a completely new user (new customer)? Please correct me if my understanding of the learning algorithm is wrong.

thanks
Arun.

 yaser 04-05-2013 10:02 PM

Re: question on netflix example

Quote:
 Originally Posted by arun.n (Post 10149) Hello everyone The way I understood the Netflix example is that, given a set of users and movies along with the ratings that the users gave to those movies, the learning algorithm learns the vectors for the profiles of users and movies. So the hypothesis is the set of vectors for the users and movies. I can see how this can be used to predict the rating that a user would give to a movie he/she didn't watch provided the user and the movie were in the original dataset. Now my question is how can this be used for a completely new user (new customer)? Please correct me if my understanding of the learning algorithm is wrong. thanks Arun.
Hi,

This is the "cold start" problem in recommender systems. There are some ways to deal with it. For instance, if users provide information when they register (demographic information or answers to initial surveys), they can be clustered with "similar" users (who have already rated some movies) based on this information and given recommendations similar to those users. As they themselves start rating movies, the system swtiches gradually to recommendations based on their ratings.

Interestingly, Netflix had a second competition that was canceled before it started that addressed the cold start problem.

 All times are GMT -7. The time now is 10:23 AM.