Hi I have a question on how the formula circulated in red from screenshot which is the derivative of z in hidden layer was calculated, as for the derivative of z in output layer it’s easy to deducted by using (dl / da) * (da / dz) like we did in previous course, but as for the hidden layer, i think i totally get lost here, can someone help explain the process to me? Thanks~
Hi @Yang_Wang1, here, we are using simple mathematics to calculate the loss while moving forward to build NN and checking the same loss again while doing the backprop. Here’s a thread that could provide you a clarity. Thanks!