In this section, we'll be developing a music generation model. We'll be using RNNs for that, and using the LSTM neuron model for the same. An RNN is different from a simple artificial neural network (ANN) in a very significant way—it allows the reuse of input between layers.
While, in an ANN, we expect input values that enter the neural network to move forward and then produce error-based feedback to be incorporated into the network weights, RNNs make the input come back to the previous layers in loops several times.
The following diagram represents an RNN neuron:
From the preceding diagram, we can see that the input after passing through the activation function in the neuron splits into two parts. One part moves forward in the network toward the next layer or output, while the other part is fed back...
