Closed spencermyoung513 closed 1 year ago
One potential issue with the score is that its minimum is 0.5 (A completely overconfident model would have an AUC of 1, and a completely underconfident model would have an AUC of 0). So maybe we should account for that? It's tricky to get a perfect 0 to 1 scale
Let me know what you think of this potential implementation of the calibration score!