AUROC computation
a7b23 opened this issue · comments
In section 4.4 "Out-of-Distribution Generalization" how do you compute the AUROC scores from the EBM model? The score function is un-normalized, as opposed to the likelihood which is between 0 and 1. So how is the AUROC score computed then?
To compute AUROC, you just need to compare the relative likelihood of real and out-of-distribution samples, which you can obtain from the energy function.
The energyevalmix function in ebm_sandbox.py shows how to compute the AUROC.
Specifically, you label real data points as 1 and fake data points as 0, and use the negative energy function as the score threshold (which does not have to be between 0 and 1)
Great, thanks a lot!