Inverted Gaussian Process Optimization for Nonparametric Koopman Operator Discovery

원문 초록

The Koopman Operator Theory opens the door for application of rich linear systems theory for computationally efficient modeling and optimal control of nonlinear systems by providing a globally linear representation for complex nonlinear systems. However, methodologies for Koopman Operator discovery struggle with the dependency on the set of selected observable functions and meaningful uncertainty quantification. The primary objective of this work is to leverage Gaussian process regression (GPR) to develop a probabilistic Koopman linear model while removing the need for heuristic observable specification. In this work, we present inverted Gaussian process optimization based Koopman Operator learning (iGPK), an automatic differentiation-based approach to simultaneously learn the observable-operator combination. We show that the proposed iGPK method is robust to observation noise in the training data, while also providing good uncertainty quantification, such that the predicted distribution consistently encapsulates the ground truth, even for noisy training data.