{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

sample midterm 1 key - CSci 4511 Sample Midterm 1 Closed...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
CSci 4511 Sample Midterm 1 Closed book, notes, laptop, cell phone The following are questions i considered asking on Midterm 1. i eventually chose 16 other problems and so am publishing these as practice questions. The material here and on the exam are roughly the same - minimax, local search and lots of A*. Easier Questions 1. The distance from a node to the goal is 10. A heuristic has returned the following estimated distances. For each one, mark whether they are from a possibly admissible heuristic (T/F). a. 15 F b. 10 T c. 5 T d. 0 T e. -10 T 2. State whether the following admissible heuristics are Useful, Not Useful or Dominant. a. h1(n) = straight line distance U b. h2(n) = 0 NU c. h3(n) = max(h1, h2) D d. h4(n) = min(h1, h2) U 3. In A*, if h(n) c(n,a,goal) then f(n) < C*. This is independent of g(n) even though , f(n)=g(n)+h(n). Why? g is the past cost which we already know and thus isn't an estimate but a perfect number. It is never too low or too high, so whether f(n) is accurate depends entirely on h(n) - we know g(n)'s contribution is perfectly accurate. 4. In what sense is A* preferable to IDA*? A* searches fewer nodes than IDA* and thus has better run time performance. 5. Which algorithm we use depends on the specific needs (i.e., fast run-time performance) and properties (i.e., non-determinism) of the problem and algorithm. Name two (not counting run- time performance and non-deterministic environments). These are listed in lecture 2. They include resources memory, run time performance, development time, tuning time and power and factors deterministic, dynamic, perfect information, information accuracy, importance of optimality, discrete state, discrete actions, state space size, presence of loops, solution density, dead ends, solution type (path or goal), degree of sensitivity to starting location, whether the solution depth is known, variable action costs, test cost, relative error cost, interruptibility, context-sensitive action costs and a whole bunch of other things 6. Name three differences between breadth-first and depth-first search (other than that one searches breadth first and the other searches depth first). BFS DFS
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
optimal not optimal memory hog (exponential) memory friendly (linear) loops OK can't handle loops Normal Questions 7. What is the "no path" run-time problem? How does hierarchical path finding help? If a solution doesn't exist, A* searches the entire search space. This is slow and probably a big variance from when it finds a solution, meaning it creates performance spikes. Hierarchical path finding groups nodes into clusters and performs path finding between clusters to make sure that one cluster can reach another. If it cannot, it doesn't engage in a detailed search, allowing it to fail fast. 8. A common problem in local search is local maxima. Name two ways to combat this problem.
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}