Fig. 3.

Predictions of DMD, EDMD and DDL on trajectories of (58). a Training trajectory starts inside the domain of convergence of the linearization, i.e. $R(0)=0.8$ (see [44]). For both DDL and EDMD the order of the monomials used is $k=5$. b Same for a different training trajectory with $R(0)=0.4$ and $k=5$ c same for $R(0)=0.1$ and $k=10$ d same for $R(0)=0.1$ and $k=18$