(Roadmap — content to be written)
This chapter covers:
The convolution operation: kernel, stride, padding
Edge detection kernels (Sobel, Laplacian)
Feature maps and receptive fields
Frequency interpretation: convolution as filtering in the Fourier domain
Why convolution is the right operation for translation-invariant feature detection
Depends on: Chapter 1 (frequency/filtering), Chapter 8 (linear algebra) Builds toward: Chapter 10 (backprop through convolution), Chapter 11 (CNNs)