Convolutional Neural Networks

What this chapter does

Convolutional neural networks teach local computation with shared weights. This bounded chapter focuses on CNN vocabulary and mechanics rather than becoming a full computer vision course.

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.

01
Introduction
Why convolution is a reusable idea about local structure and shared weights.
02
Local Connectivity
Reading small neighborhoods instead of the whole input at once.
03
Convolution Kernels
Small shared weight patterns that slide across local windows.
04
Feature Maps
Outputs that record where learned detectors respond.
05
Padding and Stride
Border handling and kernel step size as shape controls.
06
Channels
Input channels, output channels, and feature maps as parallel grids.
07
Receptive Field
The original input region that can affect a feature.
08
Parameter Sharing
Reusing the same detector across positions to reduce parameters.
09
Pooling
Local summarization through max and average pooling.
10
Why Convolution Fits Grids
Why local neighborhoods make convolution useful for images and other structured grids.
11
CNN Vocabulary in Modern Deep Learning
The convolution ideas that remain useful when reading modern architectures.

Review and practice

Close the chapter deliberately.

Use the conclusion and revision notes before the chapter exercises. Hints and solutions are collected here, while lesson-level exercises reveal their own help inline.

Conclusion

What convolution contributes before residual and gated connections.

Summary and Revision Notes

A compact review of kernels, feature maps, padding, stride, channels, receptive fields, sharing, and pooling.

Exercises

Chapter-level practice for convolutional neural networks.

Hints

Low-spoiler nudges for the Chapter 14 exercises.

Solutions

Explained solutions for the Chapter 14 exercises.