Generative model

[anferico]

This sample notebook, among other things, shows how to use an ESN in generative mode, where the network, after being trained, starts being driven by its own output to generate new data.

In the example shown in the notebook, when fit() is called on the esn object, no input data gets provided, but only output data (namely, inputDataTrain) does. What if I wanted to first train the network providing both input data and output data, then feed it with an initial input value at time t=0 and finally let it being driven by its own output (thus setting its input at time t>0 as its output at time t-1) to generate new data?