Provably Reliable Large-Scale Sampling from Gaussian Processes
Not on the map. Why?
Abstract (arXiv)
When comparing approximate Gaussian process (GP) models, it can be helpful to be able to generate data from any GP. If we are interested in how approximate methods perform at scale, we may wish to generate very large synthetic datasets to evaluate them. Na"ively doing so would cost (O(n3)) flops and (O(n2)) memory to generate a size (n) sample. We demonstrate how to scale such data generation to large (n) whilst still providing guarantees that, with high probability, the sample is indistinguishable from a sample from the desired GP.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 8 of its 19 references that are in the map. First posted on arXiv 15 Nov 2022.
Builds on
8 of 19 references.
References · 8
- 2015On the Equivalence between Kernel Quadrature Rules and Random Feature Expansionscited by 136
- 2020Efficiently Sampling Functions from Gaussian Process Posteriorscited by 125
- 2020Random Features for Kernel Approximation: A Survey on Algorithms, Theory, and Beyondcited by 107
- 2015On the Error of Random Fourier Featurescited by 76
- 2015Optimal Rates for Random Fourier Featurescited by 38
- 2018Approximation beats concentration? An approximation view on inference with smooth radial kernelscited by 33
- 2020Fast Matrix Square Roots with Applications to Gaussian Processes and Bayesian Optimizationcited by 26
- 2019Fast and Accurate Gaussian Kernel Ridge Regression Using Matrix Decompositions for Preconditioningcited by 6
Built on by
Citing papers per year: 2023: 1.
1 paper.