Do we really need to calculate db2?

I am looking at the notes as well as video lecture. I just don’t get why we need to calculate db2 in this particular case. It is not used anywhere.

1 Like

@Harshawardhan_Deshpa Look closely at what your

def linear_backward(dZ, cache):

function in the ‘Building your Deep Neural Network step-by-step’ assignment is doing and how it is being used.

Remember back-prop does not occur on only one layer.

1 Like

It is used to update b2 when we run back propagation. Every “parameter” needs a gradient because all of them are “learned” values.

1 Like

ahhh! got it, so b2 is used later. Thanks. It was not used till the part that I have completed lol

1 Like