expected output is written:
gradients[“da_prev”].shape = (5, 10)
but I think the correct output should be:
gradients[“da_prev”].shape = (8, 10)
because da_prev is the shape np.dot(W.T,dt) = (8,5) (5,10) = (8,10)
Pls check
I believe this part of the assignment contains mistakes. According to the explanations, to account for concatenation, the weights for equations 19 are the first n_a, (i.e. 𝑊𝑓=𝑊𝑓[:,:𝑛𝑎] etc…). Hence, the expected dimension is [5, 10] rather than [8, 10]. It seems to be correct, but the problem is that in order to get the expected values for da_prev we have to do it the “wrong” way (i.e. without accounting for concatenation), which produces the dimension of [8, 10].
I didn’t notice the explanations for equation 19 … I noticed only the one for equation 21.
Now it works - thanks
Hi! I am neither sure, whether the other equations are right in the explanation here. For example in 3.2 equation 7 appears to me as the equation for dat, not for dot. I did not have the energy to check the other ones yet.