Figure 6
From: Echo State Networks with Self-Normalizing Activations on the Hyper-Sphere
Comparison of the network prediction the memory–nonlinearity task for \(\nu =2.5\) and \(\tau =10\). The hyper-parameters of the networks are the same used to generate Fig. 5. Here the accuracy values are \({\gamma }_{\tanh }=0.12\), \({\gamma }_{{\rm{spherical}}}=0.63\) and \({\gamma }_{{\rm{linear}}}=0.61\).
