Inductive Bias From Layerwise Structure And Backpropagation In Neural Networks: Analysis Through Simplified Models And Empirical Frameworks