Q1. Consider the neural network in Figure 25.13. Let bias values be fixed at 0, and let the weight matrices between the input and hidden, and hidden and output layers, respectively, be: W = (w1, w2,...

Extracted text: Q1. Consider the neural network in Figure 25.13. Let bias values be fixed at 0, and let the weight matrices between the input and hidden, and hidden and output layers, respectively, be: W = (w1, w2, w3) = (1, 1, –1) W = (w, w½, wý)™ = (0.5, 1, 2)" Assume that the hidden layer uses ReLU, whereas the output layer uses sigmoid activation. Assume SSE error. Answer the following questions, when the input is x = 4 and the true response is y = 0: z1 wi w2 22 ws 23 Figure 25.13. Neural network for Q1. (a) Use forward propagation to compute the predicted output. (b) What is the loss or error value? (c) Compute the net gradient vector 8º for the output layer. (d) Compute the net gradient vector &ª for the hidden layer.