Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Probabilistic Iterative Methods for Linear Systems
Jon Cockayne, Ilse C.F. Ipsen, Chris J. Oates, Tim W. Reid
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 step size chosen adaptively instead of fixed omega | probabilistic Richardson · kernel interpolation | weak calibration (q) | No clear effect |
| replace Richardson by CG as underlying method | probabilistic iterative method · kernel interpolation | weak calibration (q): q 0.49, 0.68, 0.47 (all above 0.05) | No clear effect |
| use BayesCG instead of probabilistic iterative method | BayesCG · kernel interpolation, natural and opt priors | weak calibration (q): q 0.00 and 0.00 (below 0.05) | Hurt |
| use uninformative default prior with BayesCG | BayesCG · kernel interpolation, default prior | weak calibration (q): q = 0.03 | Hurt |
| use second-order Richardson instead of first-order | probabilistic Richardson · kernel interpolation | contraction speed | No clear effect |
| initialize x1 by one Richardson step (rich) instead of iid or corr (ablation) | second-order Richardson · kernel interpolation | contraction speed | Improved |
| use CG as underlying iteration instead of Richardson | probabilistic iterative method · kernel interpolation | convergence speed | Improved |
| increase iterations m | probabilistic Richardson · kernel interpolation | spread of output distribution | Improved |
| use optimal step size instead of default 2/3 (ablation) | probabilistic Richardson · kernel interpolation | weak calibration (q): q 0.34 -> 0.52 (default init) | No clear effect |
Closest papers by meaning
- Affine Tracing: A New Paradigm for Probabilistic Linear Solvers (NeurIPS 2026 · similarity 0.88)
- Probabilistic Linear Solvers for Machine Learning (NeurIPS 2020 · similarity 0.86)
- Probabilistic Exponential Integrators (NeurIPS 2023 · similarity 0.85)
- Stable and Scalable Probabilistic Numerical Solvers for Stiff and High-Dimensional ODEs (NeurIPS 2026 · similarity 0.84)
- How Accurately Can a Gaussian Approximate Stochastic Approximation Iterates? (NeurIPS 2026 · similarity 0.82)
- Learning to Relax: Setting Solver Parameters Across a Sequence of Linear System Instances (ICLR 2024 · similarity 0.82)
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.
