Consider the following multi$-$layer perceptron neural network that we discussed in Session

$($a$)$ Suppose that the error function for the $i$ th sample of the dataset is computed as

$J^{(i)}(y^{(i)},$hat$(y{)}^{(i)})=\frac{1}{2}(y^{(i)}-$hat$(y{)}^{(i)}{)}^2.$ In our class, we compute the forward and backward

propagation for this model. Similar to that discussion, obtain the updating rule for

weights and biases of the first layer.

$($b$)$ In case that we use this model for a binary classification problem, discuss which loss

function should be use to evaluate the performance of the model. Also, compute the

updating rule for the last layer weights and biases $)$