Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Patchwork Kriging for Large-scale Gaussian Process Regression
Chiwoo Park, Daniel Apley
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 |
|---|---|---|---|
| add boundary continuity constraints (patchwork) to local GPs (ablation) | Local GP regression · Illustrative 1-d GP (K=4) | test MSE: 0.0137 -> 0.0072 | Improved |
| add boundary continuity constraints (patchwork) to local GPs (ablation) | Local GP regression · Illustrative 1-d GP (K=8) | test MSE: 0.0269 -> 0.0123 | Improved |
| add boundary continuity constraints (patchwork) to local GPs (ablation) | Local GP regression · Illustrative 1-d GP (K=16) | test MSE: 0.0594 -> 0.0141 | Improved |
| add boundary continuity constraints (patchwork) to local GPs (ablation) | Local GP regression · Illustrative 1-d GP (K=32) | test MSE: 0.1268 -> 0.0301 | Improved |
| increase number of pseudo observations B per boundary (ablation) | Patchwork kriging · synthetic-2d boundary mismatch | mismatch of neighboring local predictions: negligible when B >= 5 | Improved |
| increase B beyond 8 for medium and long range covariance (ablation) | Patchwork kriging · Simulated GP (d=100) | I-MSE, B-MSE, MSM: does not change much for B above 8 | No clear effect |
| increase number of regions K (ablation) | Patchwork kriging · Simulated stationary GP | I-MSE, B-MSE, MSM: metrics increase in K; small effect when B > 7 | Hurt |
| replace FITC with patchwork kriging | Global approximation baseline · Simulated GP, short and medium range covariance | MSE vs computation time | Improved |
| replace PIC with patchwork kriging | PIC · PROTEIN | MSE | Hurt |
| replace PIC with patchwork kriging | PIC · PROTEIN | coverage within +/-1.5 sd (%): 50.47 -> 86.53 | Improved |
| replace RBCM with patchwork kriging | RBCM · Flight Delays (800k records) | MSE: 9790.2 -> 1188.4 | Improved |
| replace PIC (M=1000) with patchwork kriging | PIC · Flight Delays (800k records) | NLPD: 5.0742 -> 4.8917 | Improved |
Closest papers by meaning
- Efficient Computation of Gaussian Process Regression for Large Spatial Data Sets by Patching Local Gaussian Processes (JMLR 2016 · similarity 0.95)
- Practical Global and Local Bounds in Gaussian Process Regression via Chaining (AAAI 2026 · similarity 0.89)
- Jump Gaussian Process Model for Estimating Piecewise Continuous Regression Functions (JMLR 2022 · similarity 0.89)
- Scalable Gaussian-process regression and variable selection using Vecchia approximations (JMLR 2022 · similarity 0.88)
- Learning from Summarized Data: Gaussian Process Regression with Sample Quasi-Likelihood (AAAI 2025 · similarity 0.87)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 2025 · similarity 0.86)
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.
