Estimating Koopman operators for nonlinear dynamical systems: a nonparametric approach

The Koopman operator provides a linear description of non-linear systems exploiting an embedding into infinite dimensional space. Dynamic Mode Decomposition and Extended are amongst the most popular finite approximations Operator. In this paper we capture their core essence as dual version same problem, them Kernel framework. To do so, leverage RKHS suitable space for learning dynamics. Learning from length data automatically approximation induced by data. Simulations comparison with standard procedures included.