Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Robustness Guarantees for Bayesian Inference with 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 |
|---|---|---|---|
| Increase training samples | ReLU GP with a two-hidden-layer architecture · MNIST feature neighborhoods | normalized maximum prediction variance | Helped |
| Increase the number of layers | ReLU GP corresponding to a fully-connected ReLU network · MNIST feature neighborhoods | normalized maximum prediction variance | Hurt |
| Increase perturbation size from γ = 0.05 to γ = 0.15 | ReLU GP robustness analysis of MNIST SIFT features · MNIST feature neighborhoods | upper bound on robustness failure probability | Mixed |
| Use more complex architectures | Variance-thresholding defenses for ReLU networks · MNIST feature neighborhoods | effectiveness of variance thresholding | Hurt |
| Rely on SIFT feature magnitude as a proxy for robustness | ReLU GP robustness analysis of MNIST features · MNIST feature neighborhoods | relationship between feature magnitude and robustness | No clear effect |
| Use a sound GP supremum bound based on Borell-TIS and Dudley's entropy integral | Bayesian inference with Gaussian process priors · Local input neighborhoods | probability bound for robustness failure | Helped |
Closest papers by meaning
- Robust and Computation-Aware Gaussian Processes (NeurIPS 2025 · similarity 0.89)
- Adversarially Robust Optimization with Gaussian Processes (NeurIPS 2018 · similarity 0.88)
- Robust Gaussian Processes via Relevance Pursuit (NeurIPS 2024 · similarity 0.87)
- Robust and Conjugate Gaussian Process Regression (ICML 2024 · similarity 0.87)
- Bayesian Optimization via Continual Variational Last Layer Training (ICLR 2025 · similarity 0.86)
- Robust Satisficing Gaussian Process Bandits Under Adversarial Attacks (NeurIPS 2025 · similarity 0.86)
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.