44. Taking p = q = 1, use feed forward to determine the value of o:

wpr*Ap + wqr*Aq = (2.5)*(1) + (1.2)*(1) = 3.7, so Ar = 1.

wps*Ap + wqs*Aq = (-1.4)*(1) + (2.1)*(1) = 0.7, so As = 1.

wro*Ar + wso*As = (-1.6)*(1) + (1.4)*(1) = -0.2, so Ao = 0.

This is the wrong answer, so use backpropagation to adjust the weights: first, compute the errors at neurons o, r, and s.

eo = target value - computed value = 1 - 0 = 1

er = eo*wro = (1)*(-1.6) = -1.6

es = eo*wso = (1)*(1.4) = 1.4

Now compute the new weights.

newwro = oldwro + eo*Ar = -1.6 + (1)*(1) = -0.6

newwso = oldwso + eo*As = 1.4 + (1)*(1) = 2.4

newwpr = oldwpr + er*Ap = 2.5 + (-1.6)*(1) = 0.9

newwps = oldwps + es*Ap = -1.4 + (1.4)*(1) = 0

newwqr = oldwqr + er*Aq = 1.2 + (-1.6)*(1) = -0.4

newwqs = oldwqs + es*Aq = 2.1 + (1.4)*(1) = 3.5

To see if this backpropagation has succeeded in adjusting the weights, compute the output o by feed forward:

newwpr*Ap + newwqr*Aq = (0.9)*(1) + (-0.4)*(1) = 0.5, so Ar = 1.

newwps*Ap + newwqs*Aq = (0)*(1) + (3.5)*(1) = 3.5, so As = 1.

newwro*Ar + newwso*As = (-0.6)*(1) + (2.4)*(1) = 1.8,, so Ao = 1.