Consider 2 models, the first model has fewer layers & more neurons per layer and the second model has more layers & few neurons per layer. How do they react to the overfitting or underfitting of the data. Considering we are feeding the same data to both of them.
Let's say there is a factor called LN to indicate layer to neuron ratio. like LN of 2:3 indicate 2 layers with 3 neurons each. So how does changing LN will affect training?
Note: I know it's mostly about tests & trying in DL. But I'm looking for some facts that we should keep in mind with respect to LN before building a DL network.
