{moderator edit - quiz text and answers removed}
Shouldn’t the answer be True for this question?
Deep networks do overfit the data resulting in lower training error. That is why we use residual networks to reduce overfitting and to avoid vanishing gradients?
I think you should listen again to what Prof Ng says about the motivations for Residual Networks. The problem is that very deep networks are frequently not trainable at all, because of problems like vanishing and exploding gradients. What makes Residual Networks interesting is that they use the skip connections to mitigate those problems and allow us to successfully train much deeper networks than we could before.
Or to put it a bit more simply: to have an overfitting problem, you first have to actually train the network, right?
Understood, thanks a lot!