This is more of a general question. But, in the week 2 exercise’s why are wo initializing the w’s to zeros when in week 3 that you shouldn’t initialize the w’s to zero because it will cause all of the neurons will be preforming the same computation?
Basically, the two examples show that initialization with zeros works for Logistic Regression but not in Deep Neural Network where we have more than one neuron. You can check this initializing with zero the example in week3 and you will see that the resulting model will not perform the expected classification.
You can get more details by checking the link below