Fig. 4
From: Modeling nonlinear oscillator networks using physics-informed hybrid reservoir computing
Ground truth trajectory training and test span division. To generate the ground truth data, each dynamical regime was evolved for a long time (\(6200~\text {s},~\Delta t=0.1~\text {s}\)). The first 1000 steps were used for training. A gap of 1000 steps is followed by 20 warm-up and test segments each composed of 100 and 2500 steps respectively, with 400 step gaps between them.
