Diffeomorphically Learning Stable Koopman Operators
This work addresses the challenge of finding meaningful finite-dimensional representations for prediction in nonlinear systems, which is important for applications in control and analysis, though it appears incremental as it builds on existing Koopman operator methods.
The authors tackled the problem of learning stable linear predictors for nonlinear dynamics by introducing Koopmanizing Flows, a supervised continuous-time framework that learns a diffeomorphic mapping and linear dynamics simultaneously, achieving superior performance on the LASA handwriting benchmark compared to state-of-the-art methods.
System representations inspired by the infinite-dimensional Koopman operator (generator) are increasingly considered for predictive modeling. Due to the operator's linearity, a range of nonlinear systems admit linear predictor representations - allowing for simplified prediction, analysis and control. However, finding meaningful finite-dimensional representations for prediction is difficult as it involves determining features that are both Koopman-invariant (evolve linearly under the dynamics) as well as relevant (spanning the original state) - a generally unsupervised problem. In this work, we present Koopmanizing Flows - a novel continuous-time framework for supervised learning of linear predictors for a class of nonlinear dynamics. In our model construction a latent diffeomorphically related linear system unfolds into a linear predictor through the composition with a monomial basis. The lifting, its linear dynamics and state reconstruction are learned simultaneously, while an unconstrained parameterization of Hurwitz matrices ensures asymptotic stability regardless of the operator approximation accuracy. The superior efficacy of Koopmanizing Flows is demonstrated in comparison to a state-of-the-art method on the well-known LASA handwriting benchmark.