Check image for question.

Gradient back-propagation technique is one of the fundamental algorithms for training feedforward neural networks. Using the chain rule, this algorithm calculates the gradient of the loss function at different layers of the network. In subsequent stages, computed gradients will be used to update weights with optimizers such as gradient descent or stochastic gradient descent to minimize a loss function. In this question. we want to drive an expression for the gradient of a cost function with respect to the weights and biases of a simple neural network. Consider a l-hidden layer neural network as follows 'i i l was m where a: 6 IR\" "1 is the input feature vector and y 6 IR is the network output. The networks weights are W1 6 R\" \"M and 102 6 IR\