Write your answer to three decimal.

7. Consider a neural network shown below. Consider we have a cross-entropy loss function for binary classification: L=[yln(a)+(1y)ln(1a)], where a is the probability out from the output layer activation function. Welve builtea computation graph of the network as shown below. The blue letters below are intermedlate variable labels to help you understand the connection between the network architecture graph above and the computation graphto With the same condition (y=1 ) and the learning rate T1=1/2, what is the updated weight W/21 (new)? Write your answer to three decimal places. Note: Please use the computation graph method. One can calculate the gradients directly using chain rules, but if the computation graph is not used at all, it will not score properly. Try to fill the red boxes in the computation graph. This question does not need coding and the answer can be easily obtained analytically. Hint: You may use the property of z(z)=(1) Calculate new weight using the old weight and learning learning as follows: W21WW21W21L 7. Consider a neural network shown below. Consider we have a cross-entropy loss function for binary classification: L=[yln(a)+(1y)ln(1a)], where a is the probability out from the output layer activation function. Welve builtea computation graph of the network as shown below. The blue letters below are intermedlate variable labels to help you understand the connection between the network architecture graph above and the computation graphto With the same condition (y=1 ) and the learning rate T1=1/2, what is the updated weight W/21 (new)? Write your answer to three decimal places. Note: Please use the computation graph method. One can calculate the gradients directly using chain rules, but if the computation graph is not used at all, it will not score properly. Try to fill the red boxes in the computation graph. This question does not need coding and the answer can be easily obtained analytically. Hint: You may use the property of z(z)=(1) Calculate new weight using the old weight and learning learning as follows: W21WW21W21L