Suppose that we want to build a neural network that classifies two dimensional data (i.e.. X-Ix1, x21) into two classes: diamonds and crosses. We have a set of training data that is plotted as follows: la X2 XI Draw a network that can solve this classification problem. Justify your choice of network can find on the diagram. Assume a learning rate and a sigmoidal activation function in the layers ofthe network defined in la. What are the update rules for the weights of the output layer of the network? Compute (back-propagate) the errors of the hidden layer. 1b Ic Derive the gradient decent training rule for a simple network with a single unit with output o given by the formula: where W are the weights of the network and X is the input. In your answer define explicitly the cost/error function E. You may assume, that D training examples with desired outputs (t i. 1_2..... _ D) are given. What is the difference between the Best First Search and the Beam Search algorithms? ld t 2ag OCk 0% 20% of the marks