Research map / Learning theory and algorithms / Prediction and decision losses
Uncertainty Sets for Image Classifiers using Conformal Prediction
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 |
|---|---|---|---|
| Add rank-based regularization to APS scores | APS prediction sets · ImageNet-Val, ResNet152 | size: 10.4 -> 2.11 | Helped |
| Add rank-based regularization to APS scores | APS prediction sets · ImageNet-V2, ResNet152 | size: 27.2 -> 5.69 | Helped |
| Use conformal calibration instead of uncalibrated cumulative probability thresholding | Naive prediction sets · ImageNet-Val, ResNet152 | coverage: 0.895 -> 0.900 | Helped |
| Increase λ from 0 to 0.001 (ablation) | RAPS with ResNet152 · ImageNet-Val | size-stratified coverage violation: 0.070 -> 0.041 | Helped |
| Increase λ from 0 to 0.001 (ablation) | RAPS with ResNet152 · ImageNet-Val, set-size strata | coverage: 0.97 -> 0.90 | Helped |
| Increase λ from 0 to 1 (ablation) | RAPS with ResNet152 · ImageNet-Val, images with true-label rank 7 to 10 | coverage: 0.63 -> 0.00 | Hurt |
| Increase λ from 0 to 1 (ablation) | RAPS with ResNet152 · ImageNet-Val, images with true-label rank 1 | coverage: 0.95 -> 0.98 | Helped |
| Use λ=0.2 instead of λ=0 (ablation) | RAPS with ResNet152 and k_reg=1 · ImageNet-Val, 90% target coverage | size: 11.2 -> 2.1 | Helped |
| Use LAC instead of RAPS | ResNet152 prediction sets · ImageNet-Val | size: 2.09 -> 1.76 | Helped |
| Use LAC instead of RAPS | ResNet152 prediction sets · ImageNet-Val, size-stratified coverage violation at α=10% | size-stratified coverage violation: 0.039 -> 0.156 | Hurt |
Closest papers by meaning
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.87)
- Deep Probabilistic Supervision for Image Classification (NeurIPS 2026 · similarity 0.86)
- Conformal Structured Prediction (ICLR 2025 · similarity 0.86)
- Provably Reliable Conformal Prediction Sets in the Presence of Data Poisoning (ICLR 2025 · similarity 0.86)
- Class Adaptive Conformal Training (NeurIPS 2026 · similarity 0.86)
- Direct Prediction Set Minimization via Bilevel Conformal Classifier Training (ICML 2025 · similarity 0.86)
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.