AdaFocal: Calibration-aware Adaptive Focal Loss
This addresses the calibration issue for neural network practitioners, offering a method to enhance reliability in confidence scores, though it is incremental over prior focal loss techniques.
The paper tackles the calibration problem in neural networks by proposing AdaFocal, an adaptive focal loss that adjusts the regularization parameter per sample group based on validation set confidence, resulting in improved calibration and out-of-distribution detection while maintaining accuracy across image and NLP tasks.
Much recent work has been devoted to the problem of ensuring that a neural network's confidence scores match the true probability of being correct, i.e. the calibration problem. Of note, it was found that training with focal loss leads to better calibration than cross-entropy while achieving similar level of accuracy \cite{mukhoti2020}. This success stems from focal loss regularizing the entropy of the model's prediction (controlled by the parameter $γ$), thereby reining in the model's overconfidence. Further improvement is expected if $γ$ is selected independently for each training sample (Sample-Dependent Focal Loss (FLSD-53) \cite{mukhoti2020}). However, FLSD-53 is based on heuristics and does not generalize well. In this paper, we propose a calibration-aware adaptive focal loss called AdaFocal that utilizes the calibration properties of focal (and inverse-focal) loss and adaptively modifies $γ_t$ for different groups of samples based on $γ_{t-1}$ from the previous step and the knowledge of model's under/over-confidence on the validation set. We evaluate AdaFocal on various image recognition and one NLP task, covering a wide variety of network architectures, to confirm the improvement in calibration while achieving similar levels of accuracy. Additionally, we show that models trained with AdaFocal achieve a significant boost in out-of-distribution detection.