Research Article

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

Figure 3

The plot shows the difference between the true target and the used target in each training cycle for different values of when target estimation and teacher forcing are perfect. The result is used to deduce how many extra cycles of training are needed for different values of . Note that with , the used target and the true target will be equal from the start, and only one cycle of training is needed.