Netflix Offers $1 Million Prize for Major Leap in Recommendations

The New York Times is reporting that Netflix is offering a $1 million prize for anyone who can come up with a major breakthrough in movie recommendations.

Recommendation systems, also known as collaborative filtering systems, try to predict whether a customer will like a movie, book or piece of music by comparing his or her past preferences to those of other people with similar tastes. Such systems will look at, say, the last 10 books, movies or songs a customer has rated highly and try to extrapolate an 11th.

Computer scientists say that after years of steady progress in this field, there has been a slowdown — which is what Netflix executives say prompted them to offer the problem to a wide audience for solution.

Netflix will be making 100 million customer ratings available for testing, and consulted "with privacy experts to make sure that the ratings could not be traced to individual Netflix customers."

In a surprising move, Hastings noted that Netflix will publish details of the winning solution.

You can learn more at www.NetflixPrize.com.

Comments

[Aron]

I wouldn't be suprised if an existing system under development took this contest and won, and then refused to disclose their methodology. It's quite possible that the algorithm is worth more than the prize.

I wonder: Has this data been run through Movielens already and what is it's RMSE?

[Aron]

While it sounds like a contest any Joe coder can enter, having access to something such as the Googleplex for processing power is probably an essential.

[roror]

Any idea where I can get that data?

[Mike_K]

Data is at www.netflixprize.com

[WeaselBuddha]

Only a Million? Doing it on the cheap.

[type-cast]

Nice going, Aron/Weasel. You posted the same idea twice. You must be the same person.

Netflix has failed to predict my ratings for films, whenever they tried. They need to get rid of routines that do not work and make an intelligent, evolving neural net. The system shouldn't be limited by human logic or bias. It should define its own routines and modify them adaptively. The winning system won't be designed. It will be GROWN and BRED.

[munkey]

hey, um...type-cast/netflixshrill... sorry to tell you this, but cribbing notes from the screenplay for The Terminator isn't really the best way to sound well informed on the matter.

[WeaselBuddha]

If that ain't proof positive, I don't know what is, even in a thread where he ain't being accused of being an asshole, typecast acts like one. Something Shill did all the time.

The subject bores me, how about on topic.

The Times points out other examples of prizes around comercial goals, ideally you could build a biz plan around achieving the goal - because a guy in his basement isn't likely to solve this problem, and a million ain't enough to put together the capital to base a firm on solving the problem.

I'd put money on some guys at UofMN....

[Aron]

"cribbing notes from the screenplay for The Terminator"

Perfect.

And then after they've created SkyNet, which is orders of magnitude more complicated, they should give it away for free because supply is infinite and prices are arbitrary.