[2] (6 points) For nn1, let wih(i=1,2,h=1,2,3,4) denote the weight between the input variable xi and hidden node h, and let wh(h=1,2,3,4) denote the weight between hidden node h and the output node. Assume n training samples {(xl,yl)}l=1n are available, where xl=(x1l,x2l) and yl{0,1} are the input vector and the label of the lth training sample, respectively. Assume the learning algorithm finds a set of weights w that minimizes the total squared error on the training dataset: E(w)=l=1n(ylo(xl))2 Derive the partial error derivatives, i.e., whE and wihE,i=1,2,h=1,2,3,4, for nn1