Surrogate Models of Multiscale Dynamical Systems

We present a general data-driven method, continuous time echo state networks,
to produce surrogates of nonlinear multiscale systems,
which can then be used to accelerate simulation, design, control and optimization.
We showcase the ability to handle highly stiff systems, which have been shown to
cause failures in other common machine learning methods, and then showcase the ability to embed these surrogates in ModelingToolkit for control, design and optimization.