Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Training Uncertainty-Aware Classifiers with Conformalized Deep Learning
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 |
|---|---|---|---|
| train with conformal uncertainty-aware loss instead of cross entropy | CNN · CIFAR-10 corrupted images, 45000 train, full training | conditional coverage (corrupted): 0.84 -> 0.90 | Helped |
| train with conformal loss instead of cross entropy | CNN · CIFAR-10 intact images, 45000 train, full training | prediction set size (intact): 3.30 -> 1.41 | Helped |
| train with conformal loss instead of cross entropy | CNN · CIFAR-10 corrupted images, 45000 train, full training | prediction set size (corrupted): 4.43 -> 5.95 | Hurt |
| add early stopping on validation accuracy to cross entropy training | Cross Entropy · CIFAR-10 intact images, 45000 train | prediction set size (intact): 3.30 -> 1.50 | Helped |
| train with conformal loss instead of cross entropy, both with early stopping | CNN · CIFAR-10 intact images, 45000 train, ES (acc) | prediction set size (intact): 1.50 -> 1.41 | Helped |
| train with conformal loss instead of cross entropy | neural network · Credit card default, full training | conditional coverage (label 1): 0.51 -> 0.60 | Helped |
| train with conformal loss instead of hybrid loss, early stopping | neural network · Credit card default, early stopping | conditional coverage (label 1): 0.53 -> 0.52 | No clear effect |
| train with conformal loss instead of cross entropy | neural network · Credit card default, full training | classification error: 26.16 -> 33.05 | Hurt |
| conformal loss with lambda 0.2 vs cross entropy | MLP · synthetic data | conditional coverage on hard samples and set size | Helped |
Closest papers by meaning
- Class Adaptive Conformal Training (NeurIPS 2026 · similarity 0.90)
- Direct Prediction Set Minimization via Bilevel Conformal Classifier Training (ICML 2025 · similarity 0.88)
- Hierarchical Conformal Classification (NeurIPS 2026 · similarity 0.88)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.87)
- An Information Theoretic Perspective on Conformal Prediction (NeurIPS 2024 · similarity 0.87)
- Reliable Uncertainty Quantification in Machine Learning via Conformal Prediction (AAAI 2025 · similarity 0.87)
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.