1 paper
Rishabh Ranjan, Saurabh Garg, Mrigank Raman +2
Trained models are often composed with post-hoc transforms such as temperature scaling (TS), ensembling and stochastic weight averaging (SWA) to improve performance, robustness, un…