Research Article

A Novel Method for Training an Echo State Network with Feedback-Error Learning

Figure 9

In experiment 5 the network is trained on one repetition of the YMCA movement with . The plots show (a) the estimated target, (b) the used target, and (c) the ESN output after training with (b). All the plots are from epoch 5, where the used target is starting to look like the true target. Notice that the used target appears less noisy than the estimated target.
(a) Estimated target
(b) Used target
(c) ESN output after training with (b)