CSU520 Artificial Intelligence - Spring 2007 - Futrelle

CSU520 Artificial Intelligence - Spring 2007 - Exams

The quiz will be closed book, closed notes.
The Overview notes have the material in AIMA you are to read.
The Overview sample questions, one per chapter, are an excellent guide to the type of questions you will need to be able to answer. There will be around five questions on the quiz. Some will be chosen from the samples, so it behooves you to study those.
"Studying" for the quiz should consist in large part of writing out answers to each of the 27 suggested questions, with the book closed.

The topics to be covered are search and logic.
The exam will consist of questions similar to the ones on your first two assignments, but not as difficult as the more difficult one on those assignments.
More specific information will be added here as the Midterm approaches.
The Midterm will be a hybrid one. The first half of the exam will be closed book. After you hand in your answers to the first part, the remainder of the exam will be open book. This is a strategy to evaluate your performance under both conditions.

The Final Exam will be held on Friday, April 20th, 8am-10am in 104 Kariotis.
It will be a comprehensive exam, covering the entire course.
It will consist of a closed book portion followed by an open book portion.
There may be smaller numbers of questions in the two portions than in the lists below. Some may be extra-credit.
For reference, I've posted the Midterm exams here: Closed-book and Open-book.

Exercise 2.5 - Give enough details to make it clear you understand PEAS.
Exercise 3.7 - Specify the portions carefully - the initial state, goal test, successor function, and cost function.
Exercise 3.8 Search - You should have this down cold by now.
Write out the truth tables for AND, OR, and logical implication. (For implication, remember that A ⇒ B is equivalent to ¬A ∨ B)
Create a Bayesian network with reasonable CPTs for a system described for you in English.
Know the definition of conditional probability so you can prove a simple equivalence using it, as in Exercises 13.9 and 13.10.
Construct a decision tree for a simple example.
Discuss the concepts and use of training sets, test sets, the problem of overfitting and how to avoid it in building a decision tree. (Exact algorithm not required.)
A variation of the CSP question 3 of the Open-book Midterm.

Given a simple road map with distances and a table for the heuristic function h(n), show the progress of greedy best-first search or A^* search, as in figures 4.2 and 4.3 of AIMA.
For a region map I give you (not the Australian map), show how the min-conflicts and most constrained variable heuristics can be used to help solve the map coloring problem.
Convert some propositional statements to clausal form and carry through a resolution proof.
Convert some FOL statements to propositional form. You will have to introduce Skolem constants or Skolem functions. Be sure to standardize apart.
Exercise 7.7 - Create a truth table to prove a logical equivalence you will be given.
A variant of Exercise 13.6. Pay attention to bold versus non-bold P, and uppercase versus lowercase variable names. They make a difference.
Given a Bayes net, compute a conditional probability by computing the numerator and denominator from the CPTs as shown on page 495 of AIMA.
Compute the GINI values for two choices of starting attributes in a decision tree. I will give you the details of any GINI expressions you might need.