Hi, In scoring the subchallenges for #2 and #3, is there an adjustment made to the scoring logic to account for the possibility that the rmse from the covariance matrices will dominate over the rmse from the medians (because the covariance matrices have more entries (O(n^2^)) and have square terms (x~i~^2^, x~i~x~j~ etc.) ? Thanks!

Created by R S rs239
Hi thanks for noticing! we were considering to weight the different elements, however, the correlation matrix is pretty sparse. So although we get more elements from there, most of those will be close to 0. We could have come up with different weighting factors, but those are always arguable. In the end, we try to keep them as simple as possible. Since the scoring was already posted and served in the first round, we probably keep it like this, but happy to hear your opinion or suggestions. best, Attila

Scoring for subchallenges #2 and #3: median vs covariance RMSE page is loading…