Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Orthogonally Decoupled Variational Gaussian Processes
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 the orthogonally decoupled basis with natural gradient updates instead of the coupled basis with natural gradients | Variational Gaussian processes · 12 regression datasets | normalized test MAE mean: 0.290 -> 0.282 | Helped |
| Use the orthogonally decoupled basis with natural gradient updates instead of the coupled basis with natural gradients | Variational Gaussian processes · 12 regression datasets | normalized test MAE median: 0.213 -> 0.210 | Helped |
| Use the orthogonally decoupled basis instead of the coupled basis | Variational Gaussian processes · 19 classification datasets | test accuracy median (%): 94.8 -> 95.6 | Helped |
| Use natural gradient updates with the orthogonally decoupled basis instead of adaptive gradient descent (ablation) | Variational Gaussian processes · 12 regression datasets | normalized test MAE mean: 0.284 -> 0.282 | Helped |
| Use natural gradient updates with the orthogonally decoupled basis instead of adaptive gradient descent (ablation) | Variational Gaussian processes · 19 classification datasets | test accuracy median (%): 95.6 -> 93.6 | Hurt |
| Use natural gradient updates with the orthogonally decoupled basis instead of adaptive gradient descent (ablation) | Variational Gaussian processes · 19 classification datasets | test accuracy mean (%): 90.6 -> 90.3 | Hurt |
| Add an orthogonal residual mean component to the coupled basis | Variational Gaussian processes · Regression benchmarks | test performance | Helped |
| Add an orthogonal residual mean component to the coupled basis | Variational Gaussian processes · Classification benchmarks | test performance | Mixed |
| Use natural gradient updates instead of ordinary adaptive gradient updates (ablation) | Orthogonally decoupled Gaussian processes · 3droad regression illustration | convergence | Helped |
| Use natural gradient updates instead of ordinary gradient updates (ablation) | Orthogonally decoupled Gaussian processes · Classification benchmarks | test accuracy | Mixed |
| Add more gamma inducing points while keeping beta fixed (ablation) | Orthogonally decoupled Gaussian processes · Regression and classification benchmarks | test performance | Helped |
| Add more gamma inducing points while keeping beta fixed (ablation) | Hybrid and Decoupled Gaussian processes · Regression and classification benchmarks | test performance | No clear effect |
Closest papers by meaning
- Variational Inference for Gaussian Process Models with Linear Complexity (NeurIPS 2017 · similarity 0.94)
- Variational Gaussian Processes with Decoupled Conditionals (NeurIPS 2023 · similarity 0.92)
- Dual Parameterization of Sparse Variational Gaussian Processes (NeurIPS 2021 · similarity 0.91)
- New Bounds for Sparse Variational Gaussian Processes (ICML 2025 · similarity 0.90)
- Sparse Gaussian Processes: Structured Approximations and Power-EP Revisited (NeurIPS 2025 · similarity 0.90)
- Amortized Structured Stochastic Variational Inference for Gaussian Process Latent Variable Models (NeurIPS 2026 · similarity 0.89)
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.