Research map / Optimization / Implicit bias and generalization
Gaussian Process Parameter Estimation Using Mini-batch Stochastic Gradient Descent: Convergence Guarantees and Empirical Benefits
Hao Chen, Lili Zheng, Raed Al Kontar, Garvesh Raskutti
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 |
|---|---|---|---|
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · Levy | RMSE: 0.594 -> 0.264 | Improved |
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · PM2.5 | RMSE: 0.570 -> 0.286 | Improved |
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · Protein | RMSE: 0.829 -> 0.659 | Improved |
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · Query | RMSE: 0.128 -> 0.054 | Improved |
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · Energy | RMSE: 0.711 -> 0.786 | Hurt |
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · Borehole | RMSE: 0.173 -> 0.172 | No clear effect |
| use nearest-neighbor minibatches instead of uniform minibatches (ablation) | sgGP · Levy (noise variance estimate) | learned noise std (true 0.174): 0.549 -> 0.200 | Improved |
| train GP with minibatch SGD (nn) instead of exact GP with full-data inference | sgGP (nn) vs EGP · PM2.5 | training time (min): 372.88 -> 4.24 | Improved |
| train GP with minibatch SGD (nn) instead of exact GP with full-data inference | sgGP (nn) vs EGP · Protein | training time (min): 453.40 -> 2.63 | Improved |
| train GP with minibatch SGD (nn) instead of exact GP | sgGP (nn) vs EGP · Energy | RMSE: 0.802 -> 0.786 | Improved |
| use sgGP (nn) instead of SVGP | sgGP (nn) vs SVGP · Borehole | training time (min): 1380.86 -> 65.19 | Improved |
| use sgGP (nn) instead of Vecchia | sgGP (nn) vs Vecchia · Protein | RMSE: 0.597 -> 0.659 | Hurt |
Closest papers by meaning
- Stochastic Gradient Descent in Correlated Settings: A Study on Gaussian Processes (NeurIPS 2020 · similarity 0.96)
- Determinantal point processes based on orthogonal polynomials for sampling minibatches in SGD (NeurIPS 2021 · similarity 0.93)
- Sampling from Gaussian Process Posteriors using Stochastic Gradient Descent (NeurIPS 2023 · similarity 0.91)
- A Variational Analysis of Stochastic Gradient Algorithms (ICML 2016 · similarity 0.90)
- Painless Stochastic Gradient: Interpolation, Line-Search, and Convergence Rates (NeurIPS 2019 · similarity 0.90)
- Computation-Aware Gaussian Processes: Model Selection And Linear-Time Inference (NeurIPS 2024 · 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.
