Local and global sparse Gaussian process approximations

Edward Snelson, Zoubin Ghahramani
; Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics, PMLR 2:524-531, 2007.

Abstract

Gaussian process (GP) models are flexible probabilistic nonparametric models for regression, classification and other tasks. Unfortunately they suffer from computational intractability for large data sets. Over the past decade there have been many different approximations developed to reduce this cost. Most of these can be termed global approximations, in that they try to summarize all the training data via a small set of support points. A different approach is that of local regression, where many local experts account for their own part of space. In this paper we start by investigating the regimes in which these different approaches work well or fail. We then proceed to develop a new sparse GP approximation which is a combination of both the global and local approaches. Theoretically we show that it is derived as a natural extension of the framework developed by Qui onero Candela and Rasmussen [2005] for n sparse GP approximations. We demonstrate the benefits of the combined approximation on some 1D examples for illustration, and on some large real-world data sets.

Cite this Paper


BibTeX
@InProceedings{pmlr-v2-snelson07a, title = {Local and global sparse Gaussian process approximations}, author = {Edward Snelson and Zoubin Ghahramani}, booktitle = {Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics}, pages = {524--531}, year = {2007}, editor = {Marina Meila and Xiaotong Shen}, volume = {2}, series = {Proceedings of Machine Learning Research}, address = {San Juan, Puerto Rico}, month = {21--24 Mar}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v2/snelson07a/snelson07a.pdf}, url = {http://proceedings.mlr.press/v2/snelson07a.html}, abstract = {Gaussian process (GP) models are flexible probabilistic nonparametric models for regression, classification and other tasks. Unfortunately they suffer from computational intractability for large data sets. Over the past decade there have been many different approximations developed to reduce this cost. Most of these can be termed global approximations, in that they try to summarize all the training data via a small set of support points. A different approach is that of local regression, where many local experts account for their own part of space. In this paper we start by investigating the regimes in which these different approaches work well or fail. We then proceed to develop a new sparse GP approximation which is a combination of both the global and local approaches. Theoretically we show that it is derived as a natural extension of the framework developed by Qui onero Candela and Rasmussen [2005] for n sparse GP approximations. We demonstrate the benefits of the combined approximation on some 1D examples for illustration, and on some large real-world data sets.} }
Endnote
%0 Conference Paper %T Local and global sparse Gaussian process approximations %A Edward Snelson %A Zoubin Ghahramani %B Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2007 %E Marina Meila %E Xiaotong Shen %F pmlr-v2-snelson07a %I PMLR %J Proceedings of Machine Learning Research %P 524--531 %U http://proceedings.mlr.press %V 2 %W PMLR %X Gaussian process (GP) models are flexible probabilistic nonparametric models for regression, classification and other tasks. Unfortunately they suffer from computational intractability for large data sets. Over the past decade there have been many different approximations developed to reduce this cost. Most of these can be termed global approximations, in that they try to summarize all the training data via a small set of support points. A different approach is that of local regression, where many local experts account for their own part of space. In this paper we start by investigating the regimes in which these different approaches work well or fail. We then proceed to develop a new sparse GP approximation which is a combination of both the global and local approaches. Theoretically we show that it is derived as a natural extension of the framework developed by Qui onero Candela and Rasmussen [2005] for n sparse GP approximations. We demonstrate the benefits of the combined approximation on some 1D examples for illustration, and on some large real-world data sets.
RIS
TY - CPAPER TI - Local and global sparse Gaussian process approximations AU - Edward Snelson AU - Zoubin Ghahramani BT - Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics PY - 2007/03/11 DA - 2007/03/11 ED - Marina Meila ED - Xiaotong Shen ID - pmlr-v2-snelson07a PB - PMLR SP - 524 DP - PMLR EP - 531 L1 - http://proceedings.mlr.press/v2/snelson07a/snelson07a.pdf UR - http://proceedings.mlr.press/v2/snelson07a.html AB - Gaussian process (GP) models are flexible probabilistic nonparametric models for regression, classification and other tasks. Unfortunately they suffer from computational intractability for large data sets. Over the past decade there have been many different approximations developed to reduce this cost. Most of these can be termed global approximations, in that they try to summarize all the training data via a small set of support points. A different approach is that of local regression, where many local experts account for their own part of space. In this paper we start by investigating the regimes in which these different approaches work well or fail. We then proceed to develop a new sparse GP approximation which is a combination of both the global and local approaches. Theoretically we show that it is derived as a natural extension of the framework developed by Qui onero Candela and Rasmussen [2005] for n sparse GP approximations. We demonstrate the benefits of the combined approximation on some 1D examples for illustration, and on some large real-world data sets. ER -
APA
Snelson, E. & Ghahramani, Z.. (2007). Local and global sparse Gaussian process approximations. Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics, in PMLR 2:524-531

Related Material