Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Additive Gaussian Processes Revisited
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 OAK instead of the strongest listed baseline | Regression models · 24 regression datasets (average) | test RMSE: 0.445 -> 0.475 | Hurt |
| Use OAK instead of SVGP | Regression models · 24 regression datasets (average) | log likelihood: -0.295 -> -0.229 | Helped |
| Use OAK instead of GBM | Classification models · 29 classification datasets (average) | accuracy: 0.870 -> 0.872 | Helped |
| Use OAK instead of GBM | Classification models · 29 classification datasets (average) | log likelihood: -0.283 -> -0.267 | Helped |
| Use OAK instead of NN | SUSY classifier · SUSY | AuC: 0.867 ± 0.002 -> 0.865 ± 0.0004 | Hurt |
| Use OAK instead of XGBoost | Churn classifier · Churn | AuC: 0.853 ± 0.008 -> 0.856 ± 0.009 | Helped |
| Replace OAK with the product kernel ∏d (1 + k̃d) (ablation) | OAK · Concrete | parsimony of interaction orders: 0.971, 0.026, 0.003 -> 0.71, 0.16, 0.13 | Hurt |
| Omit the normalizing flow and standardize inputs instead (ablation) | OAK · Baseline datasets | predictive performance | No clear effect |
| Omit the normalizing flow and standardize inputs instead (ablation) | OAK · Baseline datasets | parsimony | Hurt |
| Use the empirical measure instead of the normalizing flow (ablation) | OAK · Concrete | number of terms needed for equivalent performance | Hurt |
| Use orthogonal additive components rather than the non-orthogonal additive GP (ablation) | Additive GP on pumadyn · pumadyn | sparse-GP convergence with inducing points | Helped |
Closest papers by meaning
- On the Identifiability and Interpretability of Gaussian Process Models (NeurIPS 2023 · similarity 0.89)
- Randomly Projected Additive Gaussian Processes for Regression (ICML 2020 · similarity 0.88)
- Gaussian Process Neural Additive Models (AAAI 2024 · similarity 0.88)
- Variational Inference for Gaussian Process Models with Linear Complexity (NeurIPS 2017 · similarity 0.88)
- Log-Linear-Time Gaussian Processes Using Binary Tree Kernels (NeurIPS 2022 · similarity 0.88)
- Scalable Log Determinants for Gaussian Process Kernel Learning (NeurIPS 2017 · 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.