TY - GEN
T1 - Online Gaussian Process regression with non-Gaussian likelihood
AU - Seiferth, David
AU - Chowdhary, G.
AU - Muhlegg, M.
AU - Holzapfel, F.
N1 - Publisher Copyright:
© 2017 American Automatic Control Council (AACC).
PY - 2017/6/29
Y1 - 2017/6/29
N2 - We present a new algorithm for GP regression over data with non-Gaussian likelihood that does not require costly MCMC sampling, or variational Bayes optimization. In our method, which we term Meta-GP, we model the likelihood by another Gaussian Process point-wise in time. This approach allows for the calculation of the posterior predictive mean and variance in an analytical way pointwise in time, leading to an online inference algorithm. As a result, our method can work with streaming data, is analytically tractable, computationally efficient while being as accurate or better than Expectation Propagation, Laplace Approximation, and MCMC inference methods for non-Gaussian likelihood data.
AB - We present a new algorithm for GP regression over data with non-Gaussian likelihood that does not require costly MCMC sampling, or variational Bayes optimization. In our method, which we term Meta-GP, we model the likelihood by another Gaussian Process point-wise in time. This approach allows for the calculation of the posterior predictive mean and variance in an analytical way pointwise in time, leading to an online inference algorithm. As a result, our method can work with streaming data, is analytically tractable, computationally efficient while being as accurate or better than Expectation Propagation, Laplace Approximation, and MCMC inference methods for non-Gaussian likelihood data.
UR - http://www.scopus.com/inward/record.url?scp=85027049917&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85027049917&partnerID=8YFLogxK
U2 - 10.23919/ACC.2017.7963429
DO - 10.23919/ACC.2017.7963429
M3 - Conference contribution
AN - SCOPUS:85027049917
T3 - Proceedings of the American Control Conference
SP - 3134
EP - 3140
BT - 2017 American Control Conference, ACC 2017
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2017 American Control Conference, ACC 2017
Y2 - 24 May 2017 through 26 May 2017
ER -