The Private Test Set

Hello everyone, I hope you are enjoying this competition as much as I am.

I have one concern with the Private Test Set that I would like to kindly bring to everyone’s attention. We learned in the live Q&A that this set contains only one site: Site F.

Considering that the goal of the competition is to design a model that is generalizable to diverse buildings and climates, is it fair to decide the winners using scores on solely one site? Considering that a team could tune a model for site F and argue in their documentation that the model would excel for various test sites, how can we all ensure that the winning models would also win on a diverse test set?

Thank you,
Ryan


2 Likes