To compute the Gramm Figure 1: Schematic of an echo state network (ESN). A dynamical core called a reservoir is
driven by input signal u(t). The states of the reservoir x(t) extended by a constant 1 and combined
linearly to produce the output y(t). The reservoir consists of N nodes interconnected
with a random weight matrix W. The connectivity between the input and the reservoir nodes
is represented with a randomly generated weight matrix Win. The reservoir states and the
constant are connected to the readout layer using the weight matrix Wout. The reservoir and
the input weights are fixed after initialization, while the output weights are learned using a
regression technique.