Diagnostic Questions - The NeurIPS 2020 Education Challenge Forum

Go back to competition Back to thread list Post in this thread

> Evaluation set


Just to clarify a rule of the game: Does the evaluation set change between our submissions and the final evaluation?

At 15 submits per day we have roughly 1000 chances to overfit a model to the final evaluation set. I just want to get a better understanding of what we are building a model for (evaluation wise).


Posted by: brooksch @ Aug. 14, 2020, 3:50 p.m.

Don't you have a limit of 100 submissions total for each of the tasks?

Posted by: nirmalpatel @ Aug. 14, 2020, 7:13 p.m.

Per person, for a ten person team, still == 1,000?

(To be honest, I don't know the actual answer I am just trying to wrap my head around the science vs. the competition)

Posted by: brooksch @ Aug. 14, 2020, 7:18 p.m.

Hi there!

1. Yes, the evaluation set (reference data) in the public evaluation phase is different from that in the private evaluation phase.
2. Well, technically you can do it under the current competition setup. Do you think it is unfair for a 10 person team to have 10 times the submission opportunity?

Posted by: moonlightlane @ Aug. 29, 2020, 2:49 a.m.
Post in this thread