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

Table 1

The table summarizes the experiment details, including the value of 𝛽 (𝛽=1 means the original method), the number of cycles per epoch, and the number of repetitions of the YMCA movement constituting the training sequence. In all the experiments the ESN was trained for 10 epochs with decreasing feedback gain.

Experiment number	𝛽	# cycles per epoch	# rep. movement
Exp. 1	1	1	1
Exp. 2	1	1	5
Exp. 3	1	1	10
Exp. 4	0.3	2	1
Exp. 5	0.1	3	1
Exp. 6	0.05	10	1