Research map / Learning theory and algorithms / Prediction and decision losses
SpeedCP: Fast Kernel-based Conditional Conformal Prediction
Yating Liu, Yeo Jin Jung, Zixuan Wu, Sowon Jeong, Claire Donnat
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 binary-search RKHS refits with the lambda/S solution-path algorithm | CondCP (Gibbs et al.) · Synthetic admixture, linear regression | computation time (s): 1332.67 -> 22.05 | Improved |
| replace binary-search RKHS refits with the solution-path algorithm | CondCP (Gibbs et al.) · QM9 | computation time (s): 1531.15 -> 31.061 | Improved |
| use SpeedCP exact path solutions instead of CondCP approximate solver | CondCP · Synthetic admixture, neural network predictor | prediction set size: 2.126 -> 0.804 | Improved |
| use SpeedCP instead of CondCP | CondCP · ESOL | prediction set size: 2.683 -> 1.789 | Improved |
| use SpeedCP instead of SplitCP | SplitCP · QM9 and ESOL | prediction set size: QM9 1.122 vs 1.135 | No clear effect |
| add linear term for predicted label to intercept-only SpeedCP (ablation) | SpeedCP(1) · Brain tumor MRI | tumor-group coverage: 0.914 -> 0.901 | Improved |
| add predicted-label linear term (ablation) | SpeedCP · Brain tumor MRI | prediction set size: 0.262 -> 0.282 | Hurt |
| tune hyperparameters by cross-validation on the calibration set instead of a held-out split (ablation) | SpeedCP · Synthetic admixture | coverage | No clear effect |
| condition on low-rank latent embeddings instead of raw covariates | RKHS conformal prediction · MRI 256-dim features | conditional coverage and set size | Improved |
| use SpeedCP instead of PCP | PCP · Brain tumor MRI | conditional coverage stability: healthy coverage 0.945 -> 0.902 | Mixed |
Closest papers by meaning
- Scalable and adaptive prediction bands with kernel sum-of-squares (NeurIPS 2025 · similarity 0.90)
- Co-optimization for Adaptive Conformal Prediction (NeurIPS 2026 · similarity 0.90)
- Optimal transport-based conformal prediction (ICML 2025 · similarity 0.89)
- A Kernel Nonconformity Score for Multivariate Conformal Prediction (NeurIPS 2026 · similarity 0.89)
- Conformal Prediction with Learned Features (ICML 2024 · similarity 0.89)
- Leave-One-Out Stable Conformal Prediction (ICLR 2025 · similarity 0.89)
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.
