Q3. (6) Differentiate between three layered perceptron and single layered perceptron, for it to be useful in performing the backpropagation algorithm. (4 marks) Describe all the steps to perform backpropagation algorithm (8 marks) Design a two layer, 1-2-1 neural network which will approximate the following trigonometric function, g(P) = 1+ sin (-p) for 25p52. The transfer function for the first layer is given as, f (1) (n) = whereas the transfer function for the second layer is linear. The following are the initial values: 1 = (1+e-n) weight, w(1)(0) = (-0:22]: w(0) = [0.03 -0.2] biases, b(1)(0) = -0.44]; b?(0) = [0.35] initial input, a = p = 1 learning rate, = 0.12 In the design it is required to produce the architecture of the network complete with the necessary labels, forward calculations, sensitivities at all layers and finally the weight and bias updates. You are only required to perform up to the first iteration. (13 marks)