this is an interesting attempt at explainability, but that's all.

The backward pass of "Shallowed" Deep network would likely result in a useless network with today's training methods.