Research map / Learning theory and algorithms / Prediction and decision losses
Approximating Full Conformal Prediction at Scale via Influence Functions
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 |
|---|---|---|---|
| Approximate full CP scores with influence functions (ACP, deleted) instead of using split conformal prediction | ACP vs SCP · MNIST, CIFAR-10, and US Census | efficiency AUC | Helped |
| Use ACP rather than RAPS | ACP · MNIST and CIFAR-10, and most US Census model settings | efficiency AUC | Helped |
| Use ACP rather than CV+ | ACP · MNIST, CIFAR-10, and US Census | efficiency AUC | Helped |
| Use RAPS instead of ACP | Logistic regression · US Census | efficiency AUC: .276 vs .183 | Helped |
| Use the deleted rather than ordinary nonconformity scheme (ablation) | ACP · MNIST, CIFAR-10, and US Census | prediction set efficiency | Helped |
| Use direct rather than indirect influence approximation (ablation) | ACP · Synthetic data | nonconformity-score approximation error | Helped |
| Increase the training-set size | ACP · Synthetic data | approximation error to full CP: <10^{-3} apart for 10^{3} training points | Helped |
| Increase regularization strength | ACP · Synthetic data | approximation error to full CP | Helped |
| Use ACP rather than full CP | ACP · Synthetic data with 1000 training points and 100 test points | runtime: approximately 5 days vs less than 1 hour | Helped |
Closest papers by meaning
- Optimal transport-based conformal prediction (ICML 2025 · similarity 0.89)
- Conformal Prediction via Regression-as-Classification (ICLR 2024 · similarity 0.89)
- Scalable and adaptive prediction bands with kernel sum-of-squares (NeurIPS 2025 · similarity 0.89)
- Backward Conformal Prediction (NeurIPS 2025 · similarity 0.88)
- Co-optimization for Adaptive Conformal Prediction (NeurIPS 2026 · similarity 0.88)
- Conformal Prediction with Learned Features (ICML 2024 · similarity 0.88)
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.