In this lesson
- It starts with a question. One neuron can draw a single dividing line through the data. What does a second layer add?
- The idea you will find. Early layers find parts; later layers find things. — In an image network the first layers respond to edges, the next to corners and textures, the ones after that to whole objects. Nobody designs that progression — it falls out of training. That is what "deep" learning means, and why depth mattered more than any single clever trick.
- Length. 5 short steps. Progress saves on this phone.
Part of Neurons & Networks
Inside the model: how a few numbers added up become a yes or a no.