Attended Temperature Scaling: A Practical Approach for Calibrating Deep Neural Networks.

Recently, Deep Neural Networks (DNNs) have been achieving impressive results on wide range of tasks. However, they suffer from being overconfident. In decision-making applications such as autonomous driving or medical diagnosing, the confidence of deep networks plays an important role to bring the trust and reliability to the system. To calibrate the deep networks many probabilistic and measure-based approaches are proposed. Temperature Scaling (TS) is a state-of-the-art measure-based calibration method which has low time and memory complexity as well as effectiveness. In this paper, we study TS functionality and show it does not work properly when the validation set that TS uses for calibration has small size or contains noisy-labeled samples. TS also cannot calibrate highly accurate networks as well as none highly accurate ones. Accordingly, we propose Attended Temperature Scaling (ATS) which preserves the advantages of TS while improves calibration in aforementioned challenging situations. We provide theoretical justifications for ATS and assess its effectiveness on wide range of deep models and datasets. We also compare the calibration results of TS and ATS on skin lesion detection application as a practical problem where well-calibrated system can play important role in making a decision.