Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Exact Gaussian Processes on a Million Data Points
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 exact GP inference instead of SGPR/SVGP | UCI regression datasets · PoleTele, Bike, Kin40K, Protein, KeggDirected, KEGGU, 3DRoad, Buzz, HouseElectric | RMSE | Helped |
| Use exact GP inference instead of SGPR/SVGP | UCI regression datasets · CTslice and Song | RMSE | Mixed |
| Increase exact GP training data from a subsample to more of the available training set (ablation) | Exact GP (BBMM) · KEGGU, 3DRoad, Song | RMSE | Helped |
| Train on less than a quarter of the training data rather than the full set (ablation) | Exact GP (BBMM) · KEGGU, 3DRoad, Song | RMSE | Helped |
| Increase the number of inducing points (ablation) | SGPR and SVGP · Bike and Protein | RMSE | No clear effect |
| Use a looser PCG training tolerance, up to ε = 1 (ablation) | Exact GP (BBMM) · UCI regression datasets | test performance | No clear effect |
| Use a loose solve tolerance for prediction-time linear solves (ablation) | Exact GP (BBMM) · UCI regression datasets | predictive performance | Hurt |
| Pretrain hyperparameters on 10,000 points, then fine-tune on the full dataset (ablation) | Exact GP (BBMM) · UCI regression datasets | RMSE | No clear effect |
| Use a larger partial pivoted-Cholesky preconditioner, up to rank 100 (ablation) | Exact GP (BBMM) · Large datasets | training time | Helped |
| Add GPUs for distributed matrix-vector products (ablation) | Exact GP (BBMM) · KEGGU, 3DRoad, Song, Buzz | training speedup | Helped |
Closest papers by meaning
- Scalable Exact Inference in Multi-Output Gaussian Processes (ICML 2020 · similarity 0.91)
- Computation-Aware Gaussian Processes: Model Selection And Linear-Time Inference (NeurIPS 2024 · similarity 0.90)
- Sparse Gaussian Processes: Structured Approximations and Power-EP Revisited (NeurIPS 2025 · similarity 0.88)
- Scalable Derivative Gaussian Processes via Exact Gradient Reduction (NeurIPS 2026 · similarity 0.88)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 2025 · similarity 0.88)
- Scalable Gaussian Processes with Latent Kronecker Structure (ICML 2025 · similarity 0.88)
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.