Q3 (8 marks) Consider a network with 10 inputs to a hidden layer of 3 nodes, and an output layer of 2 nodes. The network is fuily connected. a) (1 marks) Sketch the network. b) (2 marks) How many weights (w) and how many biases (b) are there? c) (4 marks) Suppose the training data consists of 1000 inputs, that is 1000 vectors containing 10 numbers. We will employ stochastic gradient descent to train the network, where each mini-batch contains 20 inputs. We will train the network in 3 epochs. Consider the first step in training the network: input 1 data vector with the correct output, and calculate the quadratic cost function based on a randomly chosen set of weights and biases. Now change one weight by a small amount and re-calculate the cost function. Using the stochastic gradient descent method, how many times does the cost function need to be calculated to enable the first step down the gradient? d) (1 mark) How many times does the cost function get computed in each epoch and in the full training program? To complete the rest of this assignment, you will need access to anaconda with keras (all free) Set up Python with Anaconda: hittps://machinelearningmasterv.com/setup-pvthon-environment: Q3 (8 marks) Consider a network with 10 inputs to a hidden layer of 3 nodes, and an output layer of 2 nodes. The network is fuily connected. a) (1 marks) Sketch the network. b) (2 marks) How many weights (w) and how many biases (b) are there? c) (4 marks) Suppose the training data consists of 1000 inputs, that is 1000 vectors containing 10 numbers. We will employ stochastic gradient descent to train the network, where each mini-batch contains 20 inputs. We will train the network in 3 epochs. Consider the first step in training the network: input 1 data vector with the correct output, and calculate the quadratic cost function based on a randomly chosen set of weights and biases. Now change one weight by a small amount and re-calculate the cost function. Using the stochastic gradient descent method, how many times does the cost function need to be calculated to enable the first step down the gradient? d) (1 mark) How many times does the cost function get computed in each epoch and in the full training program? To complete the rest of this assignment, you will need access to anaconda with keras (all free) Set up Python with Anaconda: hittps://machinelearningmasterv.com/setup-pvthon-environment