Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration
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.
Cite
Text
Tomani et al. "Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration." Proceedings of the European Conference on Computer Vision (ECCV), 2022. doi:10.1007/978-3-031-19778-9_32Markdown
[Tomani et al. "Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration." Proceedings of the European Conference on Computer Vision (ECCV), 2022.](https://mlanthology.org/eccv/2022/tomani2022eccv-parameterized/) doi:10.1007/978-3-031-19778-9_32BibTeX
@inproceedings{tomani2022eccv-parameterized,
title = {{Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration}},
author = {Tomani, Christian and Cremers, Daniel and Buettner, Florian},
booktitle = {Proceedings of the European Conference on Computer Vision (ECCV)},
year = {2022},
doi = {10.1007/978-3-031-19778-9_32},
url = {https://mlanthology.org/eccv/2022/tomani2022eccv-parameterized/}
}