Think Together · Start with a Short Story
The girl first builds small parts with simple blocks, then joins them into more complex shapes. Why might many layers of calculation be useful?
Lay Out the Process
- Simple inputPixels and other data
- Combine in the middleForm representations
- Transform furtherRepresent complex relationships
- Task outputMake a prediction
Many-layer networks can combine and transform information step by step to represent complex input-output relationships.
In image models, some representations may relate to edges, textures, and more complex shapes. But not every network has layers that match fixed concepts imagined by people.
Adding layers also makes training and computation harder, so the model needs a suitable structure and data.
Now You Try
If you add layers to a model but do not fix incorrect labels, will the data problem solve itself?
Ready? Open this to see the explanation
No. Incorrect targets can still lead the model to learn the wrong thing. Check both the structure and the data.
One Idea to Take Away
Many-layer networks can combine and transform information step by step to represent complex input-output relationships.
Try covering the diagram above and explaining each step to someone in your family. If one step is hard to explain, return to that card and look again.