Gaussian Process Kernels for Pattern Discovery and Extrapolation
Abstract
Gaussian processes are rich distributions over functions, which provide a Bayesian nonparametric approach to smoothing and interpolation. We introduce simple closed form kernels that can be used with Gaussian processes to discover patterns and enable extrapolation. These kernels are derived by modelling a spectral density -- the Fourier transform of a kernel -- with a Gaussian mixture. The proposed kernels support a broad class of stationary covariances, but Gaussian process inference remains simple and analytic. We demonstrate the proposed kernels by discovering patterns and performing long range extrapolation on synthetic examples, as well as atmospheric CO2 trends and airline passenger data. We also show that we can reconstruct standard covariances within our framework.
Cite
@article{arxiv.1302.4245,
title = {Gaussian Process Kernels for Pattern Discovery and Extrapolation},
author = {Andrew Gordon Wilson and Ryan Prescott Adams},
journal= {arXiv preprint arXiv:1302.4245},
year = {2014}
}
Comments
10 pages, 5 figures, 1 table. Minor edits and titled changed from "Gaussian Process Covariance Kernels for Pattern Discovery and Extrapolation" to "Gaussian Process Kernels for Pattern Discovery and Extrapolation". Appears at the International Conference on Machine Learning (ICML), JMLR W&CP 28(3):1067-1075, 2013