More netflix-style competitions?

Posted: May 11th, 2009 | Author: Alex | Filed under: Uncategorized | No Comments »

In The Million Dollar Programming Prize, Robert M. Bell, Jim Bennett, Yehuda Koren and Chris Volinsky give an account of one team’s experiences participating at the Netflix Prize competition.

The authors are pointing out some interesting findings towards the end of the article.

Now that the confetti has settled, we have a chance to look back on our work and to ask what this experience tells us. First, Netflix has incorporated our discoveries into an improved version of its algorithm, which is now being tested. Second, researchers are benefiting from the data set that the competition made available, and not just because it is orders of magnitude larger than previous data sets. It is also qualitatively better than other data sets, because Netflix gathered the information from paying customers, in a realistic setting. Third, the competition itself recruited many smart people in this line of research.

In any case, the new blood promises to quickly improve the state of the art. Such competitions have invigorated other fields. The various X Prizes that have been offered for advances in genomics, automotive design, and alternative energy have shown an excellent return: By some accounts the recent $10 million Ansari X prize, awarded for suborbital spaceflight, generated $100 million of private investment in space travel.

(From the article.)

It seems like a number of other companies could observe these lessons and offer their own versions of such competitions. Amazon.com or YouTube could make large data sets available and pose improvement of their recommendation algorithms as a challenge to the public developer community. Google could provide a data set of pages, their links, metadata, etc. and challenge developers to improve on result rankings. I am thinking a lot of companies are working on interesting, hard problems and may be able to allow outside developers and scientists to try their hands on those as well. I am wondering though how broadly the competition format can be used.

Which problems and areas of research are suitable for this approach?

---