Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
A Unifying Framework for Gaussian Process Pseudo-Point Approximations using Power Expectation Propagation
Thang D. Bui, Josiah Yan, Richard E. Turner
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 intermediate alpha=0.5 instead of VFE (ablation) | Power EP sparse GP regression · UCI regression (8 datasets) | SMSE: outperforming VFE 67% of the time | Improved |
| use intermediate alpha=0.5 instead of EP (ablation) | Power EP sparse GP regression · UCI regression (8 datasets) | SMSE: outperforming EP 78% of the time | Improved |
| move from VFE (alpha->0) to EP (alpha=1) (ablation) | Power EP sparse GP regression · UCI regression (8 datasets) | SMLL: outperforming VFE 93% of the time | Improved |
| move from alpha=0.5 to EP (alpha=1) (ablation) | Power EP sparse GP regression · UCI regression (8 datasets) | SMLL: outperforming PEP alpha=0.5 71% of the time | Improved |
| move from VFE to alpha=0.5 (ablation) | Power EP sparse GP regression · UCI regression (8 datasets) | SMLL: outperforms VFE 93% of the time | Improved |
| move from EP to VFE (ablation) | Power EP sparse GP regression · UCI regression (8 datasets) | SMSE: VFE is better than EP 64% of the time | Improved |
| increase number of pseudo-points M (ablation) | Power EP sparse GP · synthetic regression (M 5 to 500) | SMSE and SMLL | Improved |
| use alpha=0.8 or 1 with intermediate M (ablation) | Power EP sparse GP · synthetic regression | approximate NLML | Hurt |
| use alpha=0.5 instead of EP (ablation) | Power EP sparse GP classification · UCI classification (6 datasets) | NLL: 65% of the time vs. 35% | Improved |
| use larger alpha (0.8, EP) instead of smaller (ablation) | Power EP sparse GP classification · UCI classification (6 datasets) | NLL | Hurt |
Closest papers by meaning
- Sparse Gaussian Processes: Structured Approximations and Power-EP Revisited (NeurIPS 2025 · similarity 0.90)
- Streaming Sparse Gaussian Process Approximations (NeurIPS 2017 · similarity 0.89)
- State Space Gaussian Processes with Non-Gaussian Likelihood (ICML 2018 · similarity 0.89)
- Efficiently sampling functions from Gaussian process posteriors (ICML 2020 · similarity 0.88)
- A general linear-time inference method for Gaussian Processes on one dimension (JMLR 2021 · similarity 0.88)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 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.
