This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: CS570 Midterm Exam April 23, 2009 Name: Question Your Points Points 1 17 2 14 3 6 4 6 5 4 6 15 7 8 8 15 Total 85 1 1 True/False (17 points) For each of the following statements, answer True or False. Also, add a short explanation of your answer. An answer without any explanation will get zero points. A. (3 points) Bestfirst search is a greedier algorithm than A* search. Thus, it may find suboptimal solutions, but by expanding nodes closer to the goal sooner, it is guaranteed to find a goal sooner (after an equal or fewer number of node expansions) than A*. [True / False] Answer: False. A* with a good heuristic can be extremely fast. For instance, with h(n) = h*(n), A* goes straight to the goal via the optimal path. In contrast, bestfirst search may spend a lot of time on paths that turn out to be deadends. B. (3 points) Let h 1 and h 2 be two admissible heuristic functions. Then, max ( h 1 , . 5 * h 2 ) is also admissible. [True / False] Answer: True (assuming h 1 and h 2 are nonnegative. h 1 ≤ h * , and . 5 * h 2 ≤ h 2 ≤ h * , hence in any case max ( h 1 , . 5 * h 2 ) ≤ h * , and it is admissible. Note that if we allow h 2 to be negative, the answer would be false. I gave full credit if you answered True without an explicit assumption statement, but there needs to be a reasonable explanation of your answer (like all other True/False questions on this exam). C. Suppose you have a CSP problem, and you run arc consistency starting from the initial state (before any variables are assigned). After apply ing arc consistency, all variables have one or more possible value, and there is a variable V i whose domain D i has exactly one possible value remaining (  D i  = 1). (a) (2 points) There must be at least one solution to this CSP problem. [True / False] Answer: False. Arc consistency does not remove all impossible assignments, so we do not know if the remaining possible value corresponds to a solution or not. (b) (2 points) Any solution to this CSP problem must have the vari able V i instantiated to the value in D i . [True / False] 2 Answer: True. Since arc consistency only eliminates impossible assignments, we know that any solution must use the remaining assignments. D. The standard alphabeta pruning performs a depthfirst exploration (to a specified depth) of the game tree. (a) (2 points) Alphabeta pruning can be generalized to do a breadth first exploration of the game tree and still get the optimal answer. [True / False] Answer: False. The alphabeta pruning algorithm is an optimiza tion on minmax. Minmax inherently needs to look at the game tree nodes below the current node in order to assign a value to that node. (b) (2 points) Alphabeta pruning can be generalized to do an iterative deepening exploration of the game tree and still get the optimal answer. [True / False] Yes. Iterativedeepening involves repeated deathfirst searches to increasing depths. This can be done trivially with minmax and alphabeta pruning as well, even though ID does waste some work.alphabeta pruning as well, even though ID does waste some work....
View
Full Document
 Fall '10
 AliceOh
 Artificial Intelligence, Search algorithm, Constraint satisfaction, Admissible heuristic, Consistent heuristic

Click to edit the document details