Modeling Observable Dynamics: Flow Map Learning and Delay Representation
Talk by Prof. Dr. Dongbin Xiu
Date: 22.04.26 Time: 16.30 - 18.00 Room: ETH HG G 19.2
We present a mathematical and numerical framework for modeling the dynamics of observables in complex systems, termed Flow Map Learning (FML). Rather than approximating governing equations, FML constructs discrete-time evolution operators directly from data, enabling accurate prediction of the observables even when the underlying full system is unknown. We establish that, for a broad class of systems, observable dynamics admit finite-dimensional delay representations, leading to closed evolution equations with minimal memory. This perspective provides a foundation for incorporating temporal dependence in data-driven models and fast predictions of observable dynamics. We then present various numerical examples to demonstrate the efficacy of FML for long-time prediction of observable dynamics.