Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Extending Temperature Scaling with Homogenizing Maps
Christopher Qian, Feng Liang, Jason Adams
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| replace temperature scaling with softplus maps (SP-TS, SP-1, SP-U) | temperature scaling · CIFAR-100, five CNNs | ECE (10 bins) | Improved |
| enforce homogenization (slope constraint) on the softplus map (ablation) | SP-U · CIFAR-100 OOD detection (SVHN, LSUN) | AUC | Improved |
| use SP-TS (more homogenizing than TS) instead of TS | temperature scaling · CIFAR-100 OOD detection (SVHN, LSUN) | AUC | No clear effect |
| use SP-TS instead of SP-1 (greater homogenization) (ablation) | SP-1 · CIFAR-100 OOD detection | AUC: 1.122 -> 0.682 | Improved |
| add free shift parameter u (SP-U) to softplus map | SP-1 · CIFAR-100 ECE | ECE (10 bins) | No clear effect |
| add learned logit shift w^T z to SP-U (ablation) | SP-U · CIFAR-100 SVHN | avg % entropy decrease: 0.597 -> 0.651 | Hurt |
| add learned logit shift w^T z (ablation) | SP-TS, SP-1, SP-U · CIFAR-100 | ECE (10 bins) | Improved |
| add learned logit shift w^T z (ablation) | SP-U · CIFAR-100 OOD detection | AUC | Hurt |
| use SP-TS instead of TS | temperature scaling · CIFAR-100 | underconfidence | Hurt |
| aggregate 200 dropout predictions (Bayesian setting) instead of point prediction | recalibrated CNNs · CIFAR-100 | calibration and OOD AUC | Improved |
| use SP-TS instead of TS in the Bayesian setting | temperature scaling · CIFAR-100 dropout classifiers | ECE (10 bins) | Improved |
| use Approach 3 instead of Approach 1 for SP-TS | SP-TS · CIFAR-100 Bayesian | AUC | Improved |
Closest papers by meaning
- On the Limitations of Temperature Scaling for Distributions with Overlaps (ICLR 2024 · similarity 0.87)
- HoTS: Homophily-Aware Temperature Scaling for Graph Neural Network Calibration (NeurIPS 2026 · similarity 0.86)
- On Temperature Scaling and Conformal Prediction of Deep Classifiers (ICML 2025 · similarity 0.86)
- Consistency-Guided Temperature Scaling Using Style and Content Information for Out-of-Domain Calibration (AAAI 2024 · similarity 0.85)
- Power Lines: Scaling laws for weight decay and batch size in LLM pre-training (NeurIPS 2025 · similarity 0.84)
- Invertible Logits Transformation for Accuracy-Preserving Post-Hoc Uncertainty Calibration (NeurIPS 2026 · similarity 0.84)
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.
