The Elman neural network is a feedforward network in which the hidden layer, besides being connected to the output layer, forks into another identical layer, called the context layer, to which it is connected with weights equal to one. At each moment of time (each time the data is passed to the neurons of the input layer), the neurons of the context layer maintain the previous values and pass them to the respective neurons of the hidden layer. The following figure shows an Elman network scheme:
Like feedforward networks, Elman's RNNs can be trained with an algorithm called Backpropagation Through Time (BPTT), a variant ...
