Jul 08 2007

Unpredictable

Published by Lucello at 8:16 am under Geeky, In Saarbrücken

If I were a classifier, I’d be a maximum entropy model.

Most of the time, one tries to make decisions that result in the least amount of entropy and which are clear indicators of what the future might hold. Our experience is our data.

But then comes the paradox: better performance is gained by classifying only what you really know, and preserving the maximum amount of uncertainty about what you do not. Favoring the unpredictable. Admitting the outliers.

Coming to Germany was a result of this. Even the weather in Saarland tells me so. Just as in winter, the Saarland cannot seem to make up her mind about the season. One minute, she is sure it is summer, blasting us with a sultry sun. The next, it is certainly winter and she releases a tempestuous deluge. Both are short-lived. At least her conviction shows at these times and the periods of indecision are few. Gives me something to look forward to.

11 responses so far

11 Responses to “Unpredictable”

  1. # Michion 08 Jul 2007 at 11:39 am

    I don’t get it. Is the point of your post that the Saarland is female? :p

  2. # Lucelloon 08 Jul 2007 at 11:41 am

    Well, I was considering calling it a ‘he’, but I didn’t want to offend..

  3. # Lucelloon 08 Jul 2007 at 11:52 am

    To anyone who reads this: If you were a classifier, which one would you be?

  4. # becon 08 Jul 2007 at 12:22 pm

    I think I’d have to be the random baseline :P

  5. # Michion 10 Jul 2007 at 6:59 pm

    I am the naive-Bayes classifer ;)

  6. # Lucelloon 10 Jul 2007 at 11:09 pm

    Haha! Ain’t that the truth.. ;-)

  7. # moeffjuon 12 Jul 2007 at 1:08 pm

    Someone should turn that into one of these online personality quizzes.

  8. # Lucelloon 12 Jul 2007 at 1:17 pm

    So well, what does that make you?

  9. # hayat221on 12 Jul 2007 at 2:02 pm

    Perceptron….atleast it sounds sooo scientific ;)

  10. # Lucelloon 12 Jul 2007 at 2:05 pm

    You mean like, a single perceptron? So tell me, what’s your activation function.. :P

  11. # verbosuson 22 Jul 2009 at 7:16 am

    After fumbling about on Wikipedia, I think I could be the lay explanation (cheers, fellow laymen!) of a logistic regression.

Trackback URI | Comments RSS

Leave a Reply

You must be logged in to post a comment.