Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration

2021/02/24 by Christian Tomani, Tomani, Christian, Daniel Cremers +3 · 12 citations
Computer Science · Engineering · #Adversarial Robustness in Machine Learning #Anomaly Detection Techniques and Applications #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Fault Detection and Control Systems #Machine Learning (cs.LG) #cs.AI #cs.LG

paper · pdf · doi:10.48550/arxiv.2102.12182

In Proceedings of the European Conference on Computer Vision (ECCV), 2022. Code available at https://github.com/tochris/pts-uncertainty

openalex publication_date 2021/02/24 · arxiv created 2022/09/17 · arxiv updated 2022/09/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We address the problem of uncertainty calibration and introduce a novel calibration method, Parametrized Temperature Scaling (PTS). Standard deep neural networks typically yield uncalibrated predictions, which can be transformed into calibrated confidence scores using post-hoc calibration methods. In this contribution, we demonstrate that the performance of accuracy-preserving state-of-the-art post-hoc calibrators is limited by their intrinsic expressive power. We generalize temperature scaling by computing prediction-specific temperatures, parameterized by a neural network. We show with extensive experiments that our novel accuracy-preserving approach consistently outperforms existing algorithms across a large number of model architectures, datasets and metrics.

Cited by

• Uncertainty Calibration of Multi-Label Bird Sound Classifiers
• DATS: Distance-Aware Temperature Scaling for Calibrated Class-Incremental Learning
• ConfTuner: Training Large Language Models to Express Their Confidence Verbally
• Instance-Wise Monotonic Calibration by Constrained Transformation
• Rethinking Detection Calibration: A Coordinate and Direction Perspective
• Sample Margin-Aware Recalibration of Temperature Scaling
• h-calibration: Rethinking Classifier Recalibration with Probabilistic Error-Bounded Objective
• Probability Bounding: Post-Hoc Calibration via Box-Constrained Softmax
• Beyond Overconfidence: Foundation Models Redefine Calibration in Deep Neural Networks
• Adaptive Cumulative Mass Calibration with Conformal Prediction
• CRS-Triage: Confidence- and Reliability-Aware Selective Triage under Incomplete Clinical Evidence
• Enhancing Out-of-Distribution Detection with Extended Logit Normalization

Related

• Adaptive temperature scaling for Robust calibration of deep neural networks
• Adaptive Temperature Scaling for Robust Calibration of Deep Neural Networks
• Sample Margin-Aware Recalibration of Temperature Scaling
• Sample-dependent Adaptive Temperature Scaling for Improved Calibration
• Consistency-Guided Temperature Scaling Using Style and Content Information for Out-of-Domain Calibration