3 years ago

by The below write-up explains in detail with code snippets, how a RNN works from the time of initialization of weights to forward path, back propagate and update of the weight network.

Array size of 4 for 4 outputs(4 output values for each of 4 one hot encoded input elements)

Initialize the delta arrays for input weight matrix, output weight matrix and recurrent weight matrix.

The updated weight matrix values can be derived by just adding up the input and the delta values