31

I was going through an article on kaggle blogs. Repeatedly, the author mentions 'LB score' and 'LB fit') as a metric for effectiveness of machine learning (along with cross validation (CV) score).

With a research for the meaning of 'LB' I spent quite a bit of time, I realised that generally people directly refer it as LB without much background.

So my question is - What is a 'LB'?

user345394
  • 505
  • 1
  • 4
  • 8

1 Answers1

35

In the context of Kaggle, it means LeaderBoard (emphasis mine).

Emre
  • 10,481
  • 1
  • 29
  • 39
  • thanks. it seems it's just applicable for the kaggle competitions. – user345394 May 08 '17 at 05:56
  • Correct. Unimportant. – Emre May 08 '17 at 06:28
  • 2
    Worth adding that the score is *usually* just some standard metric such as log loss, and the focus on the LB score in discussion about Kaggle is around the use (and potential for over-use) of a long running public test set. – Neil Slater May 08 '17 at 06:50
  • The question is what it is - spelling out the abbreviation is not saying much about what it is. I am surprised 22 people have upvoted it ! – user702846 Oct 28 '20 at 10:44
  • It's what it sounds like: [a rankings table](https://www.dictionary.com/browse/leaderboard). – Emre Mar 09 '21 at 01:01
  • @NeilSlater Does this LB score have anything to do with evaluation metrics (say classification eval. metrics)? Or can it be translated to them? – RFAI Sep 03 '21 at 03:08
  • 2
    @RFNO Often the LB score on a Kaggle competition *is* a standard classification metric. You cannot easily translate between metrics, although usually a good score on one metric will correlate well with a good score on another. – Neil Slater Sep 03 '21 at 06:24