A.:

The main difference is that recurrent networks (RNNs) have connections going backwards or to the same layer, whereas feedforward networks (FFNs) have connections going from one layer to the next layer only. This allows RNNs to specialize in processing a series of tensors as input, whereas FFNs receive only one tensor of fixed size as input. Variations of RNNs can also produce a series of tensors, or a single tensor summarizing the input series of tensors, or receive a single input tensor and produce a series of tensors as output.