24
$\begingroup$

The voting for the site name has finished. In all, 68 people voted. There were five site names presented and each person voted for their first, second and third preference. The votes were as follows.

                     First Second Third  Weighted  Total
crossvalidated.com      29      6     4       103     39
confidenceregion.com    17     17     5        90     39
degreesoffreedom.com    15      9     5        68     29
optimummodel.com         6      3     2        26     11
modelconvergence.com     1      5     7        20     13

Just counting first preferences, CrossValidated.com is a clear winner.

If the votes are weighted (3pts for first, 2pts for second and 1pt for third), then CrossValidated.com is a clear winner.

If the total votes, regardless of preference, are counted, then it is a tie between CrossValidated.com and ConfidenceRegion.com. (Note that there was a bug in the software so that the total counts reported after you voted were incorrect.)

Consequently, the new site name will be CrossValidated.com. Thanks everyone for your participation. The next stage will be determining an appropriate logo.

For those wanting to know more about cross-validation, see Why every statistician should know about cross-validation.

| |
$\endgroup$
  • 1
    $\begingroup$ Great news. I love seeing this project moving forward. $\endgroup$ – Tal Galili Sep 28 '10 at 22:12
11
$\begingroup$

Thumbs up on crossvalidated.com. It was the catchiest, most mainstream name (even if the average person doesn't know what it means) of the lot.

| |
$\endgroup$
  • 7
    $\begingroup$ As they say, the average person has fewer than two legs. $\endgroup$ – user603 Sep 27 '10 at 12:50
  • $\begingroup$ @Kwak And 2.3 kids! $\endgroup$ – Mark C Nov 1 '10 at 4:37
-7
$\begingroup$

Please do another site name poll and include the possibility of learnedmachine.com or something more up to date. Cross-validation is only an approximation to Bayesian inference, and the five-fold method, although popular, has some serious flaws.

In particular, the size of the validation set should never be smaller than the size of the training set (see here for details). A good rule of thumb is to make the training set $n^{3/4}$ when $n$ is the overall size of the training/validation data.

That makes both the name and the site logo representative of a currently wasteful/inefficient and ineffective practice, for which I personally do not advocate. True Bayesian optimization is always more efficient; see and contribute to this demo example if you have time and/or interest.

| |
$\endgroup$
  • 3
    $\begingroup$ The name is there and familiar. It needs an extraordinarily compelling change of circumstances for us to need to change it. I wasn't a member at the time, but I read the title as far from being narrowly literal: I like to think of it as hinting strongly at the community aspect, in that we collectively support, refine, extend, and even correct each individual's expertise, experience, and judgement as applied in questions, answers and discussion. So long as there is a flag somewhere about statistics and machine learning visible to those who read carefully, the title is good. $\endgroup$ – Nick Cox Oct 15 at 7:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .