I was asked to backpropagate neural net following:
where $f$ stands for sigmoid function i.e. $f(x) = \frac{1}{1 + exp(-x)}$ and $x_1 = (1, 2)$ and $y = (1, -1)$.
My intuitive problem
My question is how is this possible that we can calulate backpropagation whereas I don't know how $y_1$ was obtained? i.e. there is no function provided how $y_1 = 1$ was received. How can I for example calculate how big impact node with $f(x)$ has on $y$ whereas I don't know how $y$ how obtained? Am I thinking correctly, or I'm missing something and we really can tell it somehow?