I have read that in deep networks you can engineer each layer for a particular purpose with regards to feature learning. I'm wondering how that is actually done and how it is trained?
In addition doesn't this conflict with the idea of deep-networks having "automatic" feature extraction?
For example consider this:
Lets say you want to detect stop signs. How would you teach a deep network to do this in a layer-wise fashion?
People write about one layer of a Deep Network does edge detection, but how?