Research map / Learning theory and algorithms / Prediction and decision losses
Guaranteed Coverage Prediction Intervals With Gaussian Process Regression
Harris Papadopoulos
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 |
|---|---|---|---|
| use normalized nonconformity measure gamma=2 instead of gamma=infinity (ablation) | GPR-CP · Artificial data (well-specified) | mean PI width 90%: 1.377 -> 1.211 | Improved |
| use gamma=2 instead of gamma=1 (ablation) | GPR-CP · Artificial data (well-specified) | mean PI width 90%: 2.360 -> 1.211 | Improved |
| use gamma=2 instead of gamma=infinity (ablation) | GPR-CP · Artificial data (outliers, unknown hyperparameters) | mean PI width 90%: 1.875 -> 1.716 | Improved |
| replace GPR variance intervals with conformal GPR-CP (gamma=2) | GPR · Artificial data (outliers, unknown hyperparameters) | mean PI width 90%: 1.834 -> 1.716 | Improved |
| replace GPR variance intervals with conformal GPR-CP for calibration | GPR · Artificial data (outliers, unknown hyperparameters) | miscoverage at 99% (%): 2.59 -> 1.03 | Improved |
| replace GPR (SE) with GPR-CP (SE, gamma=2) | GPR · Boston Housing | mean PI width 90%: 9.230 -> 8.277 | Improved |
| use GPR-CP gamma=3 instead of gamma=1 (ablation) | GPR-CP (SE) · Boston Housing | mean PI width 90%: 9.678 -> 8.140 | Improved |
| use gamma=infinity instead of gamma=3 (ablation) | GPR-CP (SE) · Boston Housing | mean PI width 90%: 8.140 -> 8.468 | Hurt |
| add GPR-CP (gamma=2) over split conformal predictive system GPR-RC | GPR-RC · Boston Housing | mean PI width 90%: 10.372 -> 8.277 | Improved |
| add GPR-CP (gamma=2) over RR-CP non-normalized measure | RR-CP · CPU Performance | mean PI width 90%: infinity -> 110.33 | Improved |
| kNNR-CP vs GPR-CP gamma=2 at 90% on Servo | GPR-CP · Servo | mean PI width 90%: 1.423 -> 1.650 | Hurt |
| change covariance function (SE, RQ, NN, Matern) (ablation) | GPR-CP · Benchmark data sets | miscoverage | No clear effect |
Closest papers by meaning
- Error Bounds for Gaussian Process Regression Under Bounded Support Noise with Applications to Safety Certification (AAAI 2025 · similarity 0.87)
- Robust and Conjugate Gaussian Process Regression (ICML 2024 · similarity 0.83)
- Conformal prediction for multi-dimensional time series by ellipsoidal sets (ICML 2024 · similarity 0.82)
- Reliable Uncertainty Quantification in Machine Learning via Conformal Prediction (AAAI 2025 · similarity 0.82)
- Distributionally Robust Active Learning for Gaussian Process Regression (ICML 2025 · similarity 0.82)
- Ensuring Class-Conditional Coverage for Pathological Workflows (Student Abstract) (AAAI 2025 · similarity 0.82)
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.
