Two Sides of Miscalibration: Identifying Over and Under-Confidence Prediction for Network Calibration

TL;DR

This paper introduces a new metric to identify over and under-confidence in neural network calibration, revealing limitations of existing methods and proposing a technique that improves calibration and failure detection.

Contribution

The paper presents a novel miscalibration score to detect both over and under-confidence, and a calibration method that addresses these issues, outperforming existing techniques.

Findings

The miscalibration score effectively identifies calibration issues.

The proposed calibration method outperforms existing techniques.

Improved failure detection and trustworthiness in neural networks.

Abstract

Proper confidence calibration of deep neural networks is essential for reliable predictions in safety-critical tasks. Miscalibration can lead to model over-confidence and/or under-confidence; i.e., the model's confidence in its prediction can be greater or less than the model's accuracy. Recent studies have highlighted the over-confidence issue by introducing calibration techniques and demonstrated success on various tasks. However, miscalibration through under-confidence has not yet to receive much attention. In this paper, we address the necessity of paying attention to the under-confidence issue. We first introduce a novel metric, a miscalibration score, to identify the overall and class-wise calibration status, including being over or under-confident. Our proposed metric reveals the pitfalls of existing calibration techniques, where they often overly calibrate the model and worsen under-confident predictions. Then we utilize the class-wise miscalibration score as a proxy to design a calibration technique that can tackle both over and under-confidence. We report extensive experiments that show our proposed methods substantially outperforming existing calibration techniques. We also validate our proposed calibration technique on an automatic failure detection task with a risk-coverage curve, reporting that our methods improve failure detection as well as trustworthiness of the model. The code are available at \url{https://github.com/AoShuang92/miscalibration_TS}.