Chapter 17 sections from Deep Learning with PyTorch.
6 items
Many learning problems involve data whose meaning depends on order.
A feedforward neural network processes inputs through a fixed sequence of layers. Once the output is produced, the computation ends. There is no memory of previous inputs.
Recurrent networks reuse the same parameters at every time step.
Recurrent neural networks were designed to process sequential data by maintaining a hidden state over time.
Standard recurrent neural networks process sequences in one direction, usually from left to right. At time step $t$, the hidden state summarizes only the past:
Recurrent neural networks were among the first deep learning architectures capable of handling variable-length sequential data.