Research map / Learning theory and algorithms / Kernels and regression theory
Compressed Gaussian Process for Manifold Regression
Rajarshi Guhaniyogi, David B. Dunson
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 random feature compression before the GP (ablation) | GP · Swiss roll n=100, p=10000, tau=0.02 | MSPE: 4.71 -> 4.09 | Improved |
| add random feature compression before the GP (ablation) | GP · Swiss roll n=5000, p=10000, tau=0.03 | MSPE x 0.1: 2.05 -> 0.56 | Improved |
| add random feature compression before the GP (ablation) | GP · Isomap faces n=648, tau=0.03 | MSPE: 0.85 -> 0.07 | Improved |
| add random feature compression before the GP (ablation) | GP · Swiss roll n=100, p=10000, tau=0.10 | MSPE: 7.09 -> 7.03 | No clear effect |
| increase feature noise tau from 0.02 to 0.10 | CGP · Swiss roll n=100, p=10000 | MSPE: 4.09 -> 7.03 | Hurt |
| increase feature dimension p from 10000 to 20000 | CGP · Swiss roll n=100, tau=0.02 | MSPE: 4.09 -> 4.43 | Hurt |
| increase feature noise tau from 0.03 to 0.10 | CGP · Swiss roll n=5000, p=10000 | MSPE x 0.1: 0.56 -> 2.18 | Hurt |
| increase feature dimension p from 10000 to 20000 | CGP · Swiss roll n=5000, tau=0.03 | MSPE x 0.1: 0.56 -> 1.17 | Hurt |
| replace compressed GP with DSL (divide-and-conquer clustering) | CGP · Swiss roll n=5000, p=10000, tau=0.03 | MSPE x 0.1: 0.56 -> 0.50 | Hurt |
| replace compressed GP with DSL | CGP · Isomap faces n=648, tau=0.03 | MSPE: 0.07 -> 0.70 | Hurt |
| use CGP instead of CRF predictive intervals | CRF · Isomap faces | 95% PI coverage | Improved |
| use CGP instead of GP for prediction intervals (ablation) | GP · Isomap faces n=648, tau=0.03 | 95% PI length: 3.06 -> 1.35 | Improved |
Closest papers by meaning
- Implicit Manifold Gaussian Process Regression (NeurIPS 2023 · similarity 0.87)
- Sketching in High-Dimensional Regression With Big Data Using Gaussian Scale Mixture Priors (JMLR 2025 · similarity 0.86)
- Wrapped Gaussian Process Regression on Riemannian Manifolds (CVPR 2018 · similarity 0.85)
- Variational Autoencoded Regression: High Dimensional Regression of Visual Data on Complex Manifold (CVPR 2017 · similarity 0.85)
- Robust Gaussian Processes via Relevance Pursuit (NeurIPS 2024 · similarity 0.84)
- Residual Deep Gaussian Processes on Manifolds (ICLR 2025 · similarity 0.84)
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.
