12

I am a somewhat new-ish member of Kaggle competitions with 1 submission. Do academics research via Kaggle challenges/competitions? If so are there any prominent names?

If not why? They provide well annotated training data-sets, and typically do not put any publication restrictions as well.

Perhaps they consider such challenges trivial. Just a query on an academic's perspective on the challenges here.

The only thing I've heard about Kaggle from an academic's perspective is from Graduate School Students at Stanford, who need to participate in a challenge as a part of their coursework (which is pretty cool).

  • I would be also very interesting to know if there are other sites/challenges apart from Kaggle! – Open the way Oct 31 '12 at 12:10
  • Are you asking whether academics submit data to Kaggle or whether they take part in Kaggle competitions? – eykanal Oct 31 '12 at 13:14
  • 3
    On a related note, this seems too discussion-oriented to me. The word "opinions" in the title tends to give it away :) – eykanal Oct 31 '12 at 13:15
  • 1
    I think there's a good question in there. The edit made minutes ago actually improves it. – F'x Oct 31 '12 at 16:30
  • 3
    A great question, but what does mean "research-worthy"? If the problems are challenging from academic point of view, if results are publishable, if it a good idea to spent some time on it to practice research-related skills...? – Piotr Migdal Oct 31 '12 at 16:43
4

The main reason "Academics" wouldn't browse Kaggle to search for research problems is that most academics already have a field of interest, and aren't interested in doing data science research for the sake of figuring out some random problem defined by whoever decided to post the contest. Even if an algorithms researcher is looking to test out a particular technique that he just devised, he would probably first apply the technique to whatever dataset he was using originally before randomly testing it out on a dataset available on Kaggle.

For what it's worth, there are hundreds of publicly available data sets to test on; if anything, the data available in Kaggle competitions is probably poor relative to the richness in those data sets.

5

One of the other uses I've noticed (besides publishing papers on a competition itself) is academics using a competition to demonstrate the application of a particular algorithm that they have developed and published. See Steffen Rendle's libFM (Factorization Machines) winning the Grockit Challenge, and the winning team of academics in the Merck Molecular Activity Challenge with their use of deep learning.

4

As far as I know, all competitions on Kaggle are on basis "who is the best". So there is no limit on skills; and actually if for someone a problem is trivial, then one can safely claim the prize (e.g. Heritage Health Prize).

From my point of view it allows both to enter interesting data, compare how one's techniques compering with others and learn other approaches (usually winners disseminate their solutions).

When it comes to the only thing in the academic world which is considered to be a serious stuff, yes, some works end up as publications, see Academic Papers - Kaggle.

Also, some competitions are research-centric (however, with no cash prizes), e.g. Eye Movements Verification and Identification Competition.

This in an official competition for BTAS 2012 (The Fifth IEEE International Conference on Biometrics: Theory, Applications and Systems, September 23-27, Washington DC, USA) and all results will be published during that conference (and of course on this web page as well).

However, there are various competitions, so I guess it's hard to make a general statement about "research-worthiness". And for all what matters the most is the result, not its 'purity'.

Out of my personal stuff:

<spam> I made an entry (a graph map of tags) for Kaggle StackExchange visualization competition. </spam>

Perhaps I'm starting in DarkWorlds, but more for fun and to practice machine learning techniques, as my field is neither machine learning, nor astrophysics.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.