21) Consider the following MLP network: X1 V2 2 a) b) 1 3 1 8.9 0.8 0.7 0.6 8.5 8.4 0.3 0.2 0.1 8 -10 -5 5 Sigmoid values: s(-5)= 0; s (-4) = .02; s(-3) = .04; s(-2) = .12; s(-1) = .27; s(0) = .5; s(5) = 1; s(1) = .98, s(3) = .96; s(2) = .88; s(1) = .73 18 In order to distribute the error for this net to all the weights, we must not only distribute the blame o directly from the final nodes 2 and 3 to w3 and w4, we must also distribute the blame from node 1 to w1 and w2. The blame for node 1 is calculated from the blame for nodes 2 and 3. Calculate the y values and 6 values. Calculate the new weights. Assume a standard error function and values: w1 = 2, w2 = -2, w3 = 4, w4 = 0; x1 = 2, x2 = 2; n = 1; d2 = 0, d3 = 1.