Analytic Long-Term Forecasting with Periodic Gaussian Processes

File Description SizeFormat 
paper.pdfAccepted version700.5 kBAdobe PDFDownload
Title: Analytic Long-Term Forecasting with Periodic Gaussian Processes
Author(s): HajiGhassemi, N
Deisenroth, MP
Item Type: Conference Paper
Abstract: Gaussian processes are a state-of-the-art method for learning models from data. Data with an underlying periodic structure appears in many areas, e.g., in climatology or robotics. It is often important to predict the long-term evolution of such a time series, and to take the inherent periodicity explicitly into account. In a Gaussian process, periodicity can be accounted for by an appropriate kernel choice. However, the standard periodic kernel does not allow for analytic long-term forecasting. To address this shortcoming, we re-parametrize the periodic kernel, which, in combination with a double approximation, allows for analytic longterm forecasting of a periodic state evolution with Gaussian processes. Our model allows for probabilistic long-term forecasting of periodic processes, which can be valuable in Bayesian decision making, optimal control, reinforcement learning, and robotics.
Publication Date: 21-Apr-2014
URI: http://hdl.handle.net/10044/1/12886
Copyright Statement: © 2014 The Authors.
Conference Name: International Conference on Artificial Intelligence and Statistics
Conference Location: Reykjavik, Iceland
Publication Status: Accepted
Start Date: 2014-04-21
Finish Date: 2014-04-25
Appears in Collections:Computing



Items in Spiral are protected by copyright, with all rights reserved, unless otherwise indicated.

Creative Commons