Self-Calibrating Vicinal Risk Minimisation for Model Calibration

Abstract

Model calibration, measuring the alignment between the prediction accuracy and model confidence, is an important metric reflecting model trustworthiness. Existing dense binary classification methods, without proper regularisation of model confidence, are prone to being over-confident. To calibrate Deep Neural Networks (DNNs), we propose a SelfCalibrating Vicinal Risk Minimisation (SCVRM) that explores the vicinity space of labeled data, where vicinal images that are farther away from labeled images adopt the groundtruth label with decreasing label confidence. We prove that in the logistic regression problem, SCVRM can be seen as a Vicinal Risk Minimisation plus a regularisation term that penalises the over-confident predictions. In practical implementation, SCVRM is approximated using Monte Carlo sampling that samples additional augmented training images and labels from the vicinal distributions. Experimental results demonstrate that SCVRM can signifi-cantly enhance model calibration for different dense classification tasks on both in-distribution and out-of-distribution data. Code is available at https://github.com/Carlisle-Liu/SCVRM.