Calibrating Deep Neural Networks using Focal Loss

TL;DR

This paper demonstrates that focal loss, combined with temperature scaling, significantly improves the calibration of deep neural networks across various datasets and architectures without sacrificing accuracy.

Contribution

It introduces the use of focal loss for better calibration of DNNs and provides a method for automatic hyperparameter selection, achieving state-of-the-art calibration results.

Findings

Focal loss improves model calibration over cross-entropy.

Combining focal loss with temperature scaling enhances calibration while maintaining accuracy.

The approach is effective across diverse datasets and network architectures.

Abstract

Miscalibration - a mismatch between a model's confidence and its correctness - of Deep Neural Networks (DNNs) makes their predictions hard to rely on. Ideally, we want networks to be accurate, calibrated and confident. We show that, as opposed to the standard cross-entropy loss, focal loss [Lin et. al., 2017] allows us to learn models that are already very well calibrated. When combined with temperature scaling, whilst preserving accuracy, it yields state-of-the-art calibrated models. We provide a thorough analysis of the factors causing miscalibration, and use the insights we glean from this to justify the empirically excellent performance of focal loss. To facilitate the use of focal loss in practice, we also provide a principled approach to automatically select the hyperparameter involved in the loss function. We perform extensive experiments on a variety of computer vision and NLP datasets, and with a wide variety of network architectures, and show that our approach achieves state-of-the-art calibration without compromising on accuracy in almost all cases. Code is available at https://github.com/torrvision/focal_calibration.