Sparse Spectrum Gaussian Process Regression
Abstract
We present a new sparse Gaussian Process (GP) model for regression. The key novel idea is to sparsify the spectral representation of the GP. This leads to a simple, practical algorithm for regression tasks. We compare the achievable trade-offs between predictive accuracy and computational requirements, and show that these are typically superior to existing state-of-the-art sparse approximations. We discuss both the weight space and function space representations, and note that the new construction implies priors over functions which are always stationary, and can approximate any covariance function in this class.
Cite
Text
Lázaro-Gredilla et al. "Sparse Spectrum Gaussian Process Regression." Journal of Machine Learning Research, 2010.Markdown
[Lázaro-Gredilla et al. "Sparse Spectrum Gaussian Process Regression." Journal of Machine Learning Research, 2010.](https://mlanthology.org/jmlr/2010/lazarogredilla2010jmlr-sparse/)BibTeX
@article{lazarogredilla2010jmlr-sparse,
title = {{Sparse Spectrum Gaussian Process Regression}},
author = {Lázaro-Gredilla, Miguel and Quiñnero-Candela, Joaquin and Rasmussen, Carl Edward and Figueiras-Vidal, Aníbal R.},
journal = {Journal of Machine Learning Research},
year = {2010},
pages = {1865-1881},
volume = {11},
url = {https://mlanthology.org/jmlr/2010/lazarogredilla2010jmlr-sparse/}
}