Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Scalable Gaussian-process regression and variable selection using Vecchia approximations
Jian Cao, Joseph Guinness, Marc G. Genton, Matthias Katzfuss
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 VGPR instead of Lasso | sparse GP regression · Slice-I | test RMSE: 0.44 -> 0.32 | Improved |
| use VGPR instead of Lasso | covariate selection · Slice-I | number of selected covariates: 792 -> 49 | Improved |
| use VGPR instead of Lasso | covariate selection · CASP-I | false positive ratio: 96% -> 0% | Improved |
| use VGPR instead of Tree | regression · Piston-D | test RMSE: 0.71 -> 0.00 | Improved |
| use VGPR instead of Lasso | regression · Temp | test RMSE: 0.28 -> 0.29 | No clear effect |
| use VGPR instead of Lasso | covariate selection · Temp | number of selected covariates: 84 -> 6 | Improved |
| add correlated artificial covariates | VGPR · CASP | test RMSE: 0.78 -> 0.75 | Improved |
| use proposed Vecchia mini-batch subsampling instead of comparison methods | SR-gradient estimator · simulated GP, n=10^4 | absolute bias and RMSE | Improved |
| add mini-batch subsampling with iterative adaptive bridge penalty | VGPR regularization path · simulated GP | computation time: reduced by more than 90% | Improved |
| use sequential warm-start addition of covariates (FWD, VGPR) instead of full-d optimization (PGPR, Fisher) | GP regression · simulated d=1,000 | RMSE | Improved |
| use VGPR instead of other GP methods | variable selection · simulation n>=5,000 | false positive ratio: zero FPos when n >= 5,000 | Improved |
| use FWD instead of VGPR | GP regression · simulation n=500 | RMSE | No clear effect |
Closest papers by meaning
- Parametric Gaussian Process Regressors (ICML 2020 · similarity 0.90)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 2025 · similarity 0.88)
- Computation-Aware Gaussian Processes: Model Selection And Linear-Time Inference (NeurIPS 2024 · similarity 0.87)
- Robust and Conjugate Gaussian Process Regression (ICML 2024 · similarity 0.87)
- Robust Gaussian Processes via Relevance Pursuit (NeurIPS 2024 · similarity 0.87)
- Learning from Summarized Data: Gaussian Process Regression with Sample Quasi-Likelihood (AAAI 2025 · similarity 0.87)
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.
