TY - JOUR
T1 - Operator-theoretic framework for forecasting nonlinear time series with kernel analog techniques
AU - Alexander, Romeo
AU - Giannakis, Dimitrios
N1 - Funding Information:
Dimitrios Giannakis acknowledges support by ONR YIP grant N00014-16-1-2649, NSF grants DMS-1521775 and 1842538, and DARPA grant HR0011-16-C-0116. Romeo Alexander was supported as a Ph.D. student from the first NSF grant and the DARPA grant. The authors would like to thank Suddhasattwa Das, Krithika Manohar, and Andrew Stuart for fruitful conversations. In addition, they would like to thank the Editor and two anonymous Reviewers for constructive comments which have led to improvements of the manuscript. Dimitrios Giannakis is grateful to the Department of Computing and Mathematical Sciences at the California Institute of Technology for hospitality and for providing a stimulating environment during a sabbatical, where part of this work was completed.
Funding Information:
Dimitrios Giannakis acknowledges support by ONR YIP grant N00014-16-1-2649 , NSF grants DMS-1521775 and 1842538 , and DARPA grant HR0011-16-C-0116 . Romeo Alexander was supported as a Ph.D. student from the first NSF grant and the DARPA grant. The authors would like to thank Suddhasattwa Das, Krithika Manohar, and Andrew Stuart for fruitful conversations. In addition, they would like to thank the Editor and two anonymous Reviewers for constructive comments which have led to improvements of the manuscript. Dimitrios Giannakis is grateful to the Department of Computing and Mathematical Sciences at the California Institute of Technology for hospitality and for providing a stimulating environment during a sabbatical, where part of this work was completed.
Publisher Copyright:
© 2020 Elsevier B.V.
PY - 2020/8
Y1 - 2020/8
N2 - Kernel analog forecasting (KAF), alternatively known as kernel principal component regression, is a kernel method used for nonparametric statistical forecasting of dynamically generated time series data. This paper synthesizes descriptions of kernel methods and Koopman operator theory in order to provide a single consistent account of KAF. The framework presented here illuminates the property of the KAF method that, under measure-preserving and ergodic dynamics, it consistently approximates the conditional expectation of observables that are acted upon by the Koopman operator of the dynamical system and are conditioned on the observed data at forecast initialization. More precisely, KAF yields optimal predictions, in the sense of minimal root mean square error with respect to the invariant measure, in the asymptotic limit of large data. The presented framework facilitates, moreover, the analysis of generalization error and quantification of uncertainty. Extensions of KAF to the construction of conditional variance and conditional probability functions, as well as to non-symmetric kernels, are also shown. Illustrations of various aspects of KAF are provided with applications to simple examples, namely a periodic flow on the circle and the chaotic Lorenz 63 system.
AB - Kernel analog forecasting (KAF), alternatively known as kernel principal component regression, is a kernel method used for nonparametric statistical forecasting of dynamically generated time series data. This paper synthesizes descriptions of kernel methods and Koopman operator theory in order to provide a single consistent account of KAF. The framework presented here illuminates the property of the KAF method that, under measure-preserving and ergodic dynamics, it consistently approximates the conditional expectation of observables that are acted upon by the Koopman operator of the dynamical system and are conditioned on the observed data at forecast initialization. More precisely, KAF yields optimal predictions, in the sense of minimal root mean square error with respect to the invariant measure, in the asymptotic limit of large data. The presented framework facilitates, moreover, the analysis of generalization error and quantification of uncertainty. Extensions of KAF to the construction of conditional variance and conditional probability functions, as well as to non-symmetric kernels, are also shown. Illustrations of various aspects of KAF are provided with applications to simple examples, namely a periodic flow on the circle and the chaotic Lorenz 63 system.
KW - Conditional expectation
KW - Kernel methods
KW - Koopman operators
KW - Statistical forecasting
UR - http://www.scopus.com/inward/record.url?scp=85084178774&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85084178774&partnerID=8YFLogxK
U2 - 10.1016/j.physd.2020.132520
DO - 10.1016/j.physd.2020.132520
M3 - Article
AN - SCOPUS:85084178774
SN - 0167-2789
VL - 409
JO - Physica D: Nonlinear Phenomena
JF - Physica D: Nonlinear Phenomena
M1 - 132520
ER -