Queries

Hey people.

Do comment on this page if you’ve got queries related to any aspect of the competition.

36 thoughts on “Queries

  1. Hi,
    It’s good that you came up with this blog, as I was looking for a way to get answers to some of my questions. I have a very basic question regarding the data set you sent. In training data, there is an unexplained label “MD”. Though I have considered it as “MP”, does it signify something else OR is “MP” the right assumption?

    • Hi Rakesh

      We’ve mailed you about this issue.

      MD in our data set refers to ‘Missing Data’. These data points are missing in the data. You may use the information in some way in your algorithm or discard them.

      Please don’t confuse it with MP or any other abbreviation.

    • Hey Shashwat

      We’d want to know what exactly you’d want on the leaderboard.
      Since multiple submissions are allowed to this contest, a feedback on how well your submission has fared would allow for training on our validation data, which is ipso facto incorrect.

      • As far as I am concerned, we don’t need to know the quantitative score a submission achieves, but a measure of where does one stand (rank etc) among all submissions so far, would certainly help us perform better. I believe without some sort of feedback mechanism permitting multiple solutions would be like shooting in the dark. One could submit N number of solutions using N different strategies and one of them might actually fit the model. That wouldn’t be actual data mining now would it ?

  2. So, just for competition’s sakes, what are the average per person penalty scores that you folks are getting ?

    • A leaderboard, in the traditional sense, refers to a list of all the submissions made to us.
      This list would also show how well you’ve performed on your submission in terms of error in prediction.

      Also, please refer to the reply posted on Shashwat’s query for a further clarification on leaderboards.

    • Hey Subhashree

      We believe we have already mailed you the data-sheets.
      However, we shall be re-posting the same on your rediffmail account.

    • I am not sure what do you mean by ‘applying the classification algorithm’. You have to come up to with a classification algorithm of your own. A good place to start may be weka. There are also third party machine learning packages for tools like R and octave.
      As far as the cost function is concerned, I have a script to evaluate the performance of a classifier created using weka schemes. Let me know if you need the same, I will be glad to help.

    • Hi Pammi
      We’d like to know the reason you’ll be needing your registration ID.
      If its for the purpose of filling up the feedback form that we’ve sent out, it’d be okay if you did not have it with you.

  3. One does not simply launch a machine learning competetion without proper user engagement.
    People are used to competetions like topcoder and kaggle whereas the aspiring-minds failed to provide even a simple leaderBoard where we can compare the submission accuracies and ranks.
    So many teams registered, thats true, but how many of them are taking this competetion seriously?
    We will see when there will be less than 10 “up to the mark” submissions.

    • Hey Dwaj

      We’re glad to hear from you. There are few things we’d like to point at and share with you –

      In order to engage with teams, we have regularly sent out mails and mobile messages to let people know of updates on our site and blog. We have also constantly been in touch with teams who have posted us their doubts via e-mail.

      On the prospects of a leaderboard, we mentioned as a reply to Shaswat’s comment [refer comment thread above] about conceptual discrepancies in providing a report on errors. As far as comparative ranking is concerned, since the initial number of submissions are low, we are engaging with the submitting-teams on a personal basis, providing them a feedback on how they can improve their design and results.

      You must appreciate that this competition is the first of its kind, here in India. Moreover, there’s a very small and rather niche crowd which is adept and would appreciate problems related to machine learning. As a consequence, we really shouldn’t expect participation in numbers we see in,say, competitive programming contests staged on platforms like Topcoder, SPOJ, Codechef etc.

      Moreover, unlike on Kaggle and other platforms which purely host competitions, we would like to rate our initial success by seeing how many engineers from this country and elsewhere we’re able to motivate and spread awareness amongst on concepts related to machine learning. We think we have taken the right first steps.

      Over this competition period, we have learned where we have lacked in our process. We are happy to learn and we strive to improve.
      We appreciate your candid feedback.

      Thanks
      ML Team
      Aspiring Minds

  4. In the year of birth column, can we compare y6 , y7 and so on, and say that candidate with year of birth y(i) is one year older than another candidate with year of birth y(i+1) ?

  5. The submission portal fails to accept pdf files. Kindly fix that as our reports are both in latex.
    Also we didn’t receive our registration ids. Our team is Data Miners.

  6. Hello, could you please extend the last date by few days?
    If yes, then thanks a load ,
    else if 25th Mar ’12 is indeed the last date, then until what time (IST or EST or UTC time) entries will be accepted.

    Please do reply,
    Thanks.

  7. As per the schedule, the results should have been out by today. Please declare the results as soon as possible by today itself

  8. The results for the Aspiring Minds Machine Learning Competition will be declared on 30th April 2012. The winners will be notified about the result on the email id provided during the registration process. Kindly remain active on our website and ML Blog for regular updates.

    Fingers Crossed!

    Best Wishes
    Aspiring Minds Team

Leave a Reply

Your email address will not be published. Required fields are marked *

*


*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>