28 May 2004 Homework 5 clarification: In the second (baseball) problem, part 3: what is meant by "Is it true that as long as you don't have BadPitching and BadBatting, your chances of losing are about the same?" is that I want you to consider each of the three cases of ~(BadPitching ^ BadBatting) (i.e, GP^GB, GP^BB, BP^BB), and tell me if the probability of losing in each of those cases is about the same as the others. 19 May 2004 Homework 5 is out.

13 May 2004 The due date for HW4 has been moved to Wednesday, 19 May.

13 May 2004 You might want to check out the newsgroup (cis.course.cis630) for some of the HW4 discussion going on there. HW4 has been updated to reflect some of the decisions made there.

10 May 2004 A minor correction has been made to Homework 3, problem 7b. See the web page for more details.

10 May 2004 Homework 4 is now out.

28 Apr 2004 In case you weren't in class on Monday, a slight change to the grading policy was approved by unanimous consent: in order to compensate for the length of the first exam, the two midterm exams will be weighted unevenly. If you score e1 on exam 1 and e2 on exam 2, the weighted average of these will be 0.6*MAX(e1,e2)+0.4*MIN(e1,e2). See the exam stats slides for more details.

28 Apr 2004 Slides for week 4 and part of week 5 are online. I've also put the rundown of the stats from the first exam online as well.

26 Apr 2004 Homework 3 is now online, due May 7th.

20 Apr 2004 Grade distributions are now online.

19 Apr 2004 Homework 2 deadline moved to Wednesday, 21 Apr 2004.

19 Apr 2004 Homework 2 FAQ's:

• In the programming assignment, should I be using the g() cost of distance + missing sample penalty for every node? No, only for the goal nodes. For every other node, you should just use the distance function. Why? Think of g() as being the cost that you would have already paid to get to that point. It's like the agent has to pay a dollar for every unit of travel. The missing sample penalty is like a tollbooth at the end: you've got to pay $x00 if you don't have x of the samples. You don't pay that cost while you're on the road, just at the end. Now, if you were clever, you could put that into the heuristic cost somehow, but that isn't required for this assignment.
• What happens if I start from the base (or in general, from any point) and I find I have to backtrack? Can you do the search with breadth-first search? The main point to keep in mind is that where you are is only part of the state of the search. So, you can revisit a place if required, but you may not be in the same search state when you do.

  ASIDE: Actually, when taken with the previous question, this question points out a deficiency in the distance metric g(). What happens if you have the pattern B->C->base->A->base? Assuming that you picked up your final sample in A, your path cost would still be 100 more than it should be, and you might pick a distance-wise non-optimal path (e.g. B->C->D->E->F->A->base might be longer but doesn't incur the penalty of entering base with not enough samples). One way to get around this is to have two "versions" of base -- one final and one non-final, the latter not incurring the penalty but also not satisifying the goal state (a sort of "passing through" state).

15 Apr 2004 Notes for the rest of week 3 are now posted.

14 Apr 2004 Notes from week 2 and week 3 (up to today) are now posted.

12 Apr 2004 Sample code for Homework 2 is now out.

7 Apr 2004 Homework 2 is now out. We haven't finished the sample code routines yet; those should be available in the next day or two.

7 Apr 2004 One student passed me an interesting link to the Sony QRIO robot.

2 Apr 2004 Homework 1 update: I have updated the C++/Java/Perl code to allow for reading from the standard input in addition to reading from a file. Sorry for the confusion. These are just called with different constructors.

31 Mar 2004 Homework 1 is out; the due date has changed from what the calendar originally said (it's now 12 April).

31 Mar 2004 I've posted the lecture notes for this week; see the calendar for details.

28 Mar 2004 For fun, here's some information about the Loebner prize that I talked about today.

• The official Loebner Prize website
• A salon.com article by John Sundman about the controversy surrounding the Loebner Prize

25 Mar 2004 Welcome to the class! Look here for future announcements. The syllabus and class calendar have been posted.