Artificial Intelligence Exam DT2001 / DT2006 Ordinarie tentamen Date: 2010-01-11 Time: 08:15-11:15 Teacher: Mathias Broxvall Phone: 301438 Aids: Calculator and/or a Swedish-English dictionary Points: The exam consists of 5 exercises with a total of 40 points. Grading: DT2001: 20 points is required for degree 3, 30 points for degree 4 and 35 points for degree 5. DT2006: 20 points is required for degree G, 32.5 points required for degree VG. Other: You may answer in either Swedish or English Use a new sheet for each exercise Motivate all answers thoroughly If anything is unclear, make reasonable assumptions and explain the assumptions.
NOTE: ALWAYS USE A NEW SHEET FOR EACH EXERCISE Exercise 1 (9 points) Answer the following questions with your own words and argue briefly for the points made. a) What is the Turing test and what is it supposed to prove. Argue briefly for the merit of the Turing test. b) What is the Chinese room experiment and what is it supposed to demonstrate. Explain the argument behind this. c) What is the Loebner prize competition. Explain how this, and recent winning competition entrants, relates to the Turing test and to the Chinese room experiment. Exercise 2 (12 points) Assume that we are doing search on the search space given by the tree below, where the goal node is the node L. The left hand side of each node is a label for each node and the right hand side a heuristic value used in exercise c. a) Explain the difference between depth first search, breadth first search and iterative deepening. Explain which ones are complete, how much time they require and how much memory they require. b) Demonstrate depth first search, breadth first search and iterative deepening by showing in which order they visit the nodes of the search tree below, stop when the goal node have been reached. c) Assume that we want to perform A*-search on the tree above. We have a cost function in which each step of the tree costs 1 and where the number in each node gives it's heuristic cost function. Demonstrate in which order the nodes of the tree above would be visited. Stop when the goal node have been reached.
Exercise 3 (6 points) a) Explain what is forward chaining in expert systems and what a proof tree is. Give an example of forward chaining and proof trees using a few rules and a few facts. b) Explain what is backward chaining in expert systems. Explain what the difference between forward and backward chaining is. Give an example of backward chaining using a few rules. c) Assume that we want to compute the probability that a patient have diabetes mellitus given that she tests positive for ketoacidocis..we know that 1% of the population at large have diabetes mellitus, and we know that 2% of the population tests positive for ketoacidocis. Furthermore, we have 100 patient records of newly diagnosed diabetics of which 50 tested positive for ketoacidocis. Compute P(Diabetes Ketoacidocis) based on these numbers. Show all the steps of the computation. Exercise 4 (6 points) What does syntactic ambiguity mean. Create, using the grammar below, a sentence that is syntactically ambigous and demonstrate that it is so using parse trees. If needed you may add extra nounds, verbs, prepositions, determinants and adjectives (but no new rules). S NP VP S NP VP PP VP V VP V NP NP N NP ADJ NP NP DET NP NP NP PP PP P NP N boy girl binoculars Homer hat P with V chases sees run DET the a ADJ young Exercise 5 (7 points) a) Explain what is a linear classifier. Give an example of a two dimensional classification problem that can be correctly classified using a linear classifier, and one that cannot be classified using such a classifier. b) Explain what a neural network is and how a perceptron functions. c) Assume that we have a classification problem of one variable and two examples. The first example has the input 0.6 should give a negative output (-1) while the second example has input 0.4 and should give a positive (+1) output. We start with a perceptron with input weight 0.0 and threshold 0.0. Demonstrate how this perceptron is trained during four iterations over the examples, with learning rates of 0.5, 0.4, 0.3 and 0.2 respectively. What would the perceptron give as a response given the input signal 0.8 after the training?