Large-Scale Gaussian Processes via Alternating Projection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Kaiwen, Wenger, Jonathan, Jones, Haydn, Pleiss, Geoff, Gardner, Jacob R.
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866907825433739264
author Wu, Kaiwen
Wenger, Jonathan
Jones, Haydn
Pleiss, Geoff
Gardner, Jacob R.
author_facet Wu, Kaiwen
Wenger, Jonathan
Jones, Haydn
Pleiss, Geoff
Gardner, Jacob R.
contents Training and inference in Gaussian processes (GPs) require solving linear systems with $n\times n$ kernel matrices. To address the prohibitive $\mathcal{O}(n^3)$ time complexity, recent work has employed fast iterative methods, like conjugate gradients (CG). However, as datasets increase in magnitude, the kernel matrices become increasingly ill-conditioned and still require $\mathcal{O}(n^2)$ space without partitioning. Thus, while CG increases the size of datasets GPs can be trained on, modern datasets reach scales beyond its applicability. In this work, we propose an iterative method which only accesses subblocks of the kernel matrix, effectively enabling mini-batching. Our algorithm, based on alternating projection, has $\mathcal{O}(n)$ per-iteration time and space complexity, solving many of the practical challenges of scaling GPs to very large datasets. Theoretically, we prove the method enjoys linear convergence. Empirically, we demonstrate its fast convergence in practice and robustness to ill-conditioning. On large-scale benchmark datasets with up to four million data points, our approach accelerates GP training and inference by speed-up factors up to $27\times$ and $72 \times$, respectively, compared to CG.
format Preprint
id arxiv_https___arxiv_org_abs_2310_17137
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Large-Scale Gaussian Processes via Alternating Projection
Wu, Kaiwen
Wenger, Jonathan
Jones, Haydn
Pleiss, Geoff
Gardner, Jacob R.
Machine Learning
Training and inference in Gaussian processes (GPs) require solving linear systems with $n\times n$ kernel matrices. To address the prohibitive $\mathcal{O}(n^3)$ time complexity, recent work has employed fast iterative methods, like conjugate gradients (CG). However, as datasets increase in magnitude, the kernel matrices become increasingly ill-conditioned and still require $\mathcal{O}(n^2)$ space without partitioning. Thus, while CG increases the size of datasets GPs can be trained on, modern datasets reach scales beyond its applicability. In this work, we propose an iterative method which only accesses subblocks of the kernel matrix, effectively enabling mini-batching. Our algorithm, based on alternating projection, has $\mathcal{O}(n)$ per-iteration time and space complexity, solving many of the practical challenges of scaling GPs to very large datasets. Theoretically, we prove the method enjoys linear convergence. Empirically, we demonstrate its fast convergence in practice and robustness to ill-conditioning. On large-scale benchmark datasets with up to four million data points, our approach accelerates GP training and inference by speed-up factors up to $27\times$ and $72 \times$, respectively, compared to CG.
title Large-Scale Gaussian Processes via Alternating Projection
topic Machine Learning
url https://arxiv.org/abs/2310.17137