search

RemyLau / fastauc

A fast implementation of auroc computation
MIT License
0 stars 0 forks source link

Integration option #5

Open RemyLau opened 2 years ago

RemyLau commented 2 years ago

https://github.com/RemyLau/fastauc/blob/979df9aef43a0743d8fd04b489cd239584160150/src/fastauc/auroc.py#L43

Currently integrating the auc using right Riemann sum. Add option to use trapezoidal rule, which is what scikit-learn uses (see source code for sklearn auc and numpy trapz)?

RemyLau commented 2 years ago

Right Riemann sum overshoots the final auc values slightly (on the order of 1e-14). Might not be a big deal but could be nice to have the option.