The following network uses ReLU, Sigmoid and Linear activations as shown below. Bias va lues at all nodes are 0. H's are outputs of hidden nodes. Y's are final outputs. (Marks 1+5+1+1=8)

1. For the given input, what will be the va lue of the squ ared error based loss function?
2. With learn ing rate= l, what wi ll be the va lue of w9, wS, w l in next i teration?(C) Find the weight for w9 in the next iteration, if we you use L2 regu l arization with value of 0.2, learn ing rate=l(D) Find the va lue of w12 with Ll regularization of 0.1, learn ing rate=l.

lnput=.2

W1-2 W5-3 X1 W9=3 H11 H2/11 Y1 W2=.5 W6=4 W 10.6 W3=2 W7=6 W11-3 X2 H7l21 Y2 H2121 W4,6 W8-5 W124.8