Research map / Label-efficient and robust learning / Recognition and long-tail
Distance-Based Learning from Errors for Confidence Calibration
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DBLE instead of vanilla training | ResNet-50 · CIFAR-100 | ECE (%): 19.1 -> 1.1 | Helped |
| use DBLE instead of vanilla training | ResNet-50 · CIFAR-100 | NLL: 1.58 -> 1.09 | Helped |
| use DBLE instead of vanilla training | ResNet-50 · Tiny-ImageNet | ECE (%): 25.2 -> 3.6 | Helped |
| use DBLE instead of vanilla training | ResNet-50 · Tiny-ImageNet | NLL: 2.95 -> 2.38 | Helped |
| add calibration learning from errors to vanilla training (ablation) | vanilla training · CIFAR-100 | ECE (%): 19.1 -> 18.3 | Helped |
| add calibration learning from errors to vanilla training (ablation) | vanilla training · Tiny-ImageNet | NLL: 2.95 -> 2.61 | Helped |
| train the confidence model using all training samples instead of misclassified samples (ablation) | DBLE · CIFAR-100 | ECE (%): 1.1 -> 18.9 | Hurt |
| train the confidence model using all training samples instead of misclassified samples (ablation) | DBLE · Tiny-ImageNet | NLL: 2.38 -> 2.87 | Hurt |
Closest papers by meaning
- On Calibration of Modern Neural Networks (ICML 2017 · similarity 0.88)
- Better Uncertainty Calibration via Proper Scores for Classification and Beyond (NeurIPS 2022 · similarity 0.87)
- Attribution-Based Confidence Metric For Deep Neural Networks (NeurIPS 2019 · similarity 0.86)
- A Benchmark Study on Calibration (ICLR 2024 · similarity 0.86)
- Improving model calibration with accuracy versus uncertainty optimization (NeurIPS 2020 · similarity 0.86)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.