Introduction to AI, Perceptrons and SUPPORT VECTOR MACHINES. b Y c a 1 . 2 . 4 3 X Given some data (x,y) they can be classified as two types (type 1 is labelled X, type two is labelled 0). Three decion boundaries are given (a, b, c). 1. Is the data linearly seperatble? 2. Could you use a perceptron (single layer network) to learn this data? 3. Which of the three decions boundaries agree with the data (i.e. which ones a or b or c separate the data )? According to each decision boundary (a, b, c, d) label each of the four points (1, 2, 3, 4) with its class (either 0 or 1). 1 2 3 4 A B C On the diagram label the a. Support vectors (how many are there)? b. Draw the best decision boundary you can (i.e. which has maximum separation of the two classes) c. Label the margin – can you measure it in millimetres. What advantage is there to using a support vector machine compared to a perceptron? (Think about a perceptron has a learning rate and initialization values, where is the final decision boundary places, what is the time complexity).