Machine Learning Philosophy – Does There Exist A Two Check Model?

If you adapt the model so that output* as input to the second group layers produce the same input* as your original input. Then I wonder if this can work as a noise filter for third model as output* to output.

A bit confusing but I think I will test it.