Long-Time Predictive Modeling of Nonlinear Dynamical Systems Using Neural Networks

Figure 21

A priori and a posteriori result of the basic model on the VDP system, training with random sampled data distribution. (a) A priori evaluation. (b) A posteriori evaluation. (c) Learning curve. (d) Stepwise error contour. Blue dot: ground truth. Orange dot: prediction. White arrow: direction of target vector of ground truth. Red arrow: direction of target vector of prediction.

(a)

(b)

(c)

(d)