## CSG120 Artificial Intelligence - Spring 2005 - Assignments

### There are four classes of assignmenets:

#### • Problems from AIMA

These are not typically programming questions, but deal more with AI concepts. See the specific assignments below.

#### • Lisp programming using PAIP

There is one AI/Lisp programming assignments. Details can be found below.

### Specific assignments and due dates

Assignment #1, AIMA, due in class, January 13th: Hand in, or email, written answers to Exercises 2.1, 2.4 and 2.12 (no Lisp code).

Assignment #3, AIMA, due in class as hardcopy, February 17th: Do as many of the problems below as you have time for. You should finish all of them before the Midterm, which will be on the 24th. I will go over the answers during class, so you'll know about them in advance of the Midterm.

• Exercise 8.6.
• Exercise 8.16.
• Exercise 9.3.
• Exercise 9.4.
• Exercise 9.9.
• Exercise 9.10 -- this is tricky, so be alert.
• Exercise 9.12.
• Exercise 9.18.

Assignment #4, AIMA, due Tuesday, April 12th, 11:59pm. This is an important assignment, since the questions are very similar to the ones you will be asked on the Final Exam. I will go over the answers to these in the second half of the last class, April 14th as your review for the Final. Hand in all your answers by the due date/time. If you find that you have to draw diagrams or use complex notation that doesn't work well for you electronically, hand in only these additional materials in class on the 14th.

• Exercise 14.1
• Exercise 16.1
• Exercise 16.2
• Exercise 16.4
• Exercise 16.10. You will need to make up some probability values to use for this exercise.
• Exercise 18.1
• Exercise 18.2
• Exercise 18.3
• Exercise 18.4
• Experiment with the restaurant domain example in Sec. 18.3.
• Split on two other attributes instead of Type and Patron and discuss the resulting trees.
• Delete some of the examples when deriving the full tree as in Fig. 18.6. Then test the tree on the ones you omitted. Discuss the successes and failures on testing on the omitted examples.
• Attempt to avoid overtraining by not insisting on perfect performance on the training set, but improving performance on unseen examples. This is easy to do if you use, say, only half the training examples.
• Include a contradictory example and discuss the problem that arises, e.g., the same example repeated but with Yes vs. No.