Knowledge MaxxingConvolutional neural network
A convolutional neural network stacks layers of convolutions and pooling to progressively build from simple local patterns, like edges, to complex ones, like whole objects.
Prerequisite: Convolution
What matters
This architecture is especially well suited to image and grid-like data, since it directly exploits spatial structure, unlike a generic fully connected network that treats every pixel independently.
Example
A convolutional neural network for photo classification might detect edges in its first layer, shapes in a middle layer, and whole objects like 'car' in a later layer.
Common misconception
A convolutional neural network is not the only architecture used for images today; vision-language and other transformer-based models are also common for image tasks.
Recall question
Why is spatial structure important for a convolutional neural network to exploit?