TUM Logo

Lazy Gaussian Process Committee for Real-Time Online Regression

A significant problem of Gaussian process (GP) is its unfavorable scaling with a large amount of data. To overcome this issue, we present a novel GP approximation scheme for online regression. Our model is based on a combination of multiple GPs with random hyperparameters. The model is trained by incrementally allocating new examples to a selected subset of GPs. The selection is carried out efficiently by optimizing a submodular function. Experiments on real-world data sets showed that our method outperforms existing online GP regression methods in both accuracy and efficiency. The applicability of the proposed method is demonstrated by the mouse-trajectory prediction in an Internet banking scenario.

Lazy Gaussian Process Committee for Real-Time Online Regression

27th AAAI Conference on Artificial Intelligence (AAAI '13)

Authors: Han Xiao and Claudia Eckert
Year/month: 2013/7
Booktitle: 27th AAAI Conference on Artificial Intelligence (AAAI '13)
Address: Washington, USA
Note: (AR: 29%)
Fulltext: aaai2.pdf

Abstract

A significant problem of Gaussian process (GP) is its unfavorable scaling with a large amount of data. To overcome this issue, we present a novel GP approximation scheme for online regression. Our model is based on a combination of multiple GPs with random hyperparameters. The model is trained by incrementally allocating new examples to a selected subset of GPs. The selection is carried out efficiently by optimizing a submodular function. Experiments on real-world data sets showed that our method outperforms existing online GP regression methods in both accuracy and efficiency. The applicability of the proposed method is demonstrated by the mouse-trajectory prediction in an Internet banking scenario.

Bibtex:

@inproceedings { hanxiao2013c,
author = { Han Xiao and Claudia Eckert},
title = { Lazy Gaussian Process Committee for Real-Time Online Regression },
year = { 2013 },
month = { July },
booktitle = { 27th AAAI Conference on Artificial Intelligence (AAAI '13) },
address = { Washington, USA },
note = { (AR: 29%) },
url = {https://www.sec.in.tum.de/i20/publications/lazy-gaussian-process-committee-for-real-time-online-regression/@@download/file/aaai2.pdf}
}