Research map / Learning theory and algorithms / Prediction and decision losses
Exact Optimization of Conformal Predictors via Incremental and Decremental 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 |
|---|---|---|---|
| Apply exact incremental/decremental updates to k-NN scores | Full k-NN CP · Synthetic classification, 100k training examples | prediction time: roughly 2 hours -> 0.63 seconds | Helped |
| Apply exact incremental/decremental updates to LS-SVM | Full LS-SVM CP · Synthetic classification, 100k training examples | prediction time: more than 24.5 hours -> 0.21 seconds | Helped |
| Reuse bootstrap samples and pretrain classifiers when possible | Random Forest CP · Synthetic classification, n=46415 | prediction time: 82 hours -> 43 hours | Helped |
| Use incremental/decremental updates for full CP instead of the ICP split | LS-SVM conformal prediction · Synthetic classification, 100k training examples | prediction time: 0.16 seconds -> 6.68 seconds | Hurt |
| Use incremental/decremental updates for full CP instead of the ICP split | KDE conformal prediction · Synthetic classification, 100k training examples | prediction time: 0.31 seconds -> 0.46 seconds | Hurt |
| Use full CP rather than ICP | Conformal prediction on MNIST · MNIST test set | fuzziness: 0.00065 ± 0.00143 -> 0.00047 ± 0.00105*; 0.05684 ± 0.07744 -> 0.04998 ± 0.07151*; 0.00098 ± 0.00174 -> 0.00066 ± 0.00125 | Helped |
| Precompute k-NN regression distances and update affected scores incrementally | Full k-NN CP regression · k-NN CP regression, 100k training examples | prediction time: 1 hour -> 9.3 seconds | Helped |
| Parallelize optimized k-NN CP across processes | Optimized k-NN CP · 100k training examples | prediction time: 3.5 hours -> 1 hour | Helped |
| Parallelize optimized k-NN CP across processes | Optimized k-NN CP · Dataset of 1000 examples | prediction time: 1.60 ± 0.27 -> 9.04 ± 0.33 | Hurt |
Closest papers by meaning
- Conformal Prediction via Regression-as-Classification (ICLR 2024 · similarity 0.88)
- Scalable and adaptive prediction bands with kernel sum-of-squares (NeurIPS 2025 · similarity 0.87)
- Class Adaptive Conformal Training (NeurIPS 2026 · similarity 0.86)
- Direct Prediction Set Minimization via Bilevel Conformal Classifier Training (ICML 2025 · similarity 0.86)
- Optimal transport-based conformal prediction (ICML 2025 · similarity 0.85)
- Boosted Conformal Prediction Intervals (NeurIPS 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.