Chapter 12

Long-Range Dependency Problems

Long dependencies, vanishing gradients, exploding gradients, memory bottlenecks, and why attention became attractive.

What this chapter does

Processing tokens in order does not guarantee that distant information remains usable. This chapter makes the limitation visible.

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

    Why long dependencies expose recurrent-model limits.

  2. 02
    Why Long Dependencies Matter

    Agreement, reference, code, and extended context often depend on distant tokens.

  3. 03
    Vanishing Gradients

    Repeated paths can make distant gradient signals very small.

  4. 04
    Exploding Gradients

    Repeated paths can make gradients unstable and too large.

  5. 05
    Memory Bottlenecks

    A fixed hidden state must compress useful past information.

  6. 06
    Why Attention Became Attractive

    Attention gives a more direct route to earlier positions.

Before moving on

  • Explain why ordered processing can still lose distant information.
  • Recognize gradient problems through time.
  • Use the limitation to motivate gated recurrence.

Where this leads

  • LSTMs and GRUs
  • From Recurrence to Attention

Chapter progress