Hello, everyone
I passed exercise 1 but I have two questions.
First, in the first function is the order in the last step “x-shortcut, x”. In the second funktion is the order in the last step “x, x_shortcut”. Why is it the other way around here?
Second, in the last exercise I become full points from the rater, but I have only a Test Accuracy from 0,74. Why is my test accuracy be so low?
Of course addition is commutative, so the order shouldn’t matter in terms of the results. But note that the order may well affect the way in which the compute graph is expressed here. That does matter in the case that they are doing a compare between the expected graph and the actual graph. The point of this architecture with the “skip layers” is that the compute graph is no longer simply connected.