C4W1 CNN back propagation

[yunxiangqin]

When calculating db by summing up dZ (optional assignment), why don’t we need to divide by number of training examples m?

[jonaslalin]

Probably because the code computes dJ/db and not dL/db as per lecture videos. I like the former notation better because it is a cleaner solution than to introduce dL/db and dL/dW. We have a recent topic where you can read more:

Typo in back prop formula (week3 and week 4)

OK, I must be making some kind of blunder here. What I proposed is to define all the partial derivatives with respect to J, instead of what you called L, which is an m-dimensional vector, when we are differentiating with respect to Z's. The point is that \frac{\partial L}{\partial Z} is of dimension m*\dim Z, which messes things up when we use them for calculations for dW and db's. It will also make things a lot more consistent. In this case, \frac{\partial J}{\partial Z^{[L]}}= \frac{1}{m}(A…