Chapter 11

Recurrent Neural Networks

Step-by-step sequence processing, hidden state, shared parameters across time, unrolling, and backpropagation through time.

What this chapter does

Recurrence is the first serious neural sequence model: one update rule is reused as a hidden state moves through the sequence.

Lessons

Read these in order.

Start with the chapter introduction, then move through the topic lessons. The order is chosen so each page can reuse ideas from the pages before it.

  1. 01
    Introduction

    RNNs as step-by-step sequence models with hidden state.

  2. 02
    Sequence Processing Step by Step

    Reading tokens in order and updating hidden state at each step.

  3. 03
    Hidden State

    The learned running summary carried through a sequence.

  4. 04
    Shared Parameters Across Time

    One recurrent update rule reused at every sequence position.

  5. 05
    Unrolling Through Time

    Drawing repeated recurrent computation as a sequence graph.

  6. 06
    Backpropagation Through Time

    Training an RNN by backpropagating through the unrolled graph.

Before moving on

  • Trace how an RNN carries information from earlier tokens.
  • Explain shared parameters across time.
  • Recognize BPTT as backpropagation through an unrolled graph.

Where this leads

  • Long-Range Dependency Problems
  • LSTMs and GRUs

Chapter progress