This page is where information about the class will be
distributed. You should check here
regularly for updates to the schedule, important announcements, and other items
of interest. You are responsible for being aware of the material on this
page.
Contents
April 22: The final exam is tomorrow, April 23^{rd}, at
1:30pm. It will be in the usual
exam rooms: 1200 and 1005. Usual
format (except 2 hours long).
April 22: As announced in class last Tuesday, solutions for PS6 are available through the assignment itself.
Apr 6: PS5s are graded. Sergej will have them at discussion and office hours today, and they’ll be at class on Tuesday. Average was 75 (but median was in the low 80s; the tail (in part due to late penalties) was long.) Some comments from the grader:
Task
1 (10pts): I gave people partial credit for a handwaving but correct answer
and full credit for an actual (correct) proof.
Task
2 (30pts): A lot of people really had problems here. Common problems were not
using the cutset method in 2.1, and getting pretty confused by 2.4 and 2.5.
Typically, errors in the beginning propagated through, so I tried to follow
people's reasoning and give partial credit for doing the right thing with the
wrong numbers.
Task
3 (20pts): A mixed bag here. A lot of people assumed the phenotypes were
equally proportional in task 3.2, which is not valid. Many people had problems
with 3.3, esp. the first two parts. When P(F) is 1/3, they could work it out,
but they got confused when p(jedi) = 0.0001, even though they're basically the
same problem.
Task
4 (15pts): Most people did fine  a few folks are a little fuzzy on explaining
away.
Task
5 (25pts): Common problems were forgetting to use the priors from task 3 or
using phenotypes as the variable values. Again, I tried to give partial credit
where possible.
Apr 3: Solutions to the second midterm are here. The mean was about a 65.
Apr 2: The version space example from class is in the handouts directory (here).
Mar 23: Midterm 2 is next Tuesday, starting at 9am (not 9:10)!! Same format as MT1; you can bring a 1page cheat sheet, and a calculator might be handy. I have reserved 1005 EECS as an overflow room.
Mar 23: Solutions to PS5 can be found here.
Mar 22: PS4 was handed back today. Solutions will be linked through the assignment as soon as possible. The average on the assignment was ~76% .
Mar 13: Prof. Durfee’s office hours on Wed 3/14 will start late, at roughly 10am. If you need to set up a separate appointment, send him email.
Mar 8: PS3 was handed back today. Mean was 76.3; median was 80.
Grader
comments: In general, it seems like
there's some folks who need to start working on these homeworks sooner &
putting a little more effort into it. Quite a few people didn't do the last
couple of questions, which hurt their grade quite a bit.
The first four tasks were each worth 10
points. Most folks did fine on these, although they may have lost a few points
for skipping steps.
The last four questions were worth 15 points
each.
Task 5 is where it started to go downhill for
a lot of folks  most people lost significant points on this one.
Task 6 was also a problem  people lost
points for not doing the CNF conversion properly or not giving enough detail in
their proofs. (e.g., just writing a series of sentences & not saying how
they got from one to the next.) Some folks also gave some informal
Englishlanguage proofs, which were not helpful.
Task 7 most folks did OK on  a lot of people
lost a few points for not listing all the answers to who was afraid of Phobos.
Task 8, again, most people did OK, although quite a few people lost some points for not doing a refutation proof. Reading the assignment carefully would have helped them.
Feb 20: Solutions to the midterm will be placed here. The mean was about a 70.
Feb 10: Examples from class are here.
Feb 8: Remember that the first midterm is on February 15^{th}, starting promptly at 9am!
Feb 8: Remember that PS3 cannot be turned in late.
Jan 30: Professor Durfee will be out of town the rest of this week. He will not hold office hours on January 31. Sergej will give the lecture on February 1.
Jan 30: The second assignment is due today. The third will be on the web shortly.
Jan 16: The second assignment (PS2) is on the web, and is due on January 30^{th}.
Jan 16: To my knowledge, override requests have been processed. Students wishing to enroll should drop from the waitlist and CRISP into the course.
Jan 8: Remember, the first assignment is due on January 16^{th}.
Jan 4: Welcome to EECS 492!
Lectures: TueThu, 9:0010:30, 1200 EECS [Durfee]
Sections:
(2) Tue 12:30  1:30, 3427 EECS [Bartold]
(3) Thu 12:30  1:30, 3427 EECS [Bartold]
(4) Fri 1:302:30, 3437 EECS [Roytman]
To contact us, email is often
best. Include "492" in message subject lines!
Course Staff
To contact the entire course staff,
use the email alias: EECS492@umich.edu.
Professor
office: 120 ATL
email: durfee@umich.edu
fax: (734)
7631260 [if you must]
Office hours:
120 ATL, Wednesdays, 910:30am, or by appointment
Email: tbartold@umich.edu
Office hours:
Mondays, 47pm, 3^{rd} Floor of the Media Union
Sergej Roytman
Email: ftit@umich.edu
Office hours:
Fridays, 2:304:30pm, 3^{rd} Floor of the Media Union
The purpose of this course is to
introduce the student to the major ideas and techniques of Artificial
Intelligence, as well as to develop an appreciation for the engineering issues
underlying the design of intelligent computational agents. The successful
student will finish the course with specific modeling and analytical skills
(e.g. search, logic, probability), knowledge of many of the most important knowledge
representation, reasoning, and machine learning schemes, and a general
understanding of AI principles and practice. The course will serve to prepare
the student for further study of AI, as well as to inform any work involving
the design of computer programs for substantial application domains.
Although earlier in the morning than
many would like, lectures are an important part of this course. While we will
generally follow the course text, lecture is an opportunity to present
additional examples, to clarify murkier parts of the text, and to cover
complementary technical material.
You will be responsible for material covered in lecture, whether or not
it aligns exactly with the text.
Weekly onehour discussion sections
will focus on developing programming and other problemsolving skills. Students
should attend the same section every week, and are encouraged to bring problems
(perhaps based on current assignments) to discuss. These smaller groups are
designed to be interactive; students are expected to participate in the
solution of problems under discussion.
Office Hours
The instructor and the GSIs will
have regularly scheduled office hours each week. You are encouraged to make use of these to discuss aspects
of the course including lecture material and the homework problems. In cases
where you cannot make office hours, contact the course staff to arrange an
appointment; don’t wait until the last minute though!
EECS
380, or equivalent experience. We assume programming experience and knowledge
of programming language concepts, and familiarity with algorithmic concepts
such as graph search and computational complexity. We will not hesitate to
employ mathematics where appropriate.
Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach, PrenticeHall, 1995.
http://ai.eecs.umich.edu/people/durfee/courses/492winter01. The course homepage is the location of
official announcements and other information. You should look at it several
times a week to remain updated. You are responsible for keeping aware of
announcements through this means. You will also find the files needed for
assignments through this homepage, as well as course handouts, syllabus
updates, and all sorts of other useful stuff.
Email directed to the course staff
may be reposted to the website (for example, for FAQs for assignments) unless
requested otherwise.
The course has a newsgroup at umich.eecs.class.492 .
The newsgroup is for student use; the instructors do not regularly look at the
newsgroup, and so information on the newsgroup does not necessarily reflect
the opinions of the course staff. For official announcements, the course
website is used. However, often you can get an answer to a question by posting
a message to the newsgroup, and someone else in the class will provide you with
an answer (but do not post code or problem solutions!). To read the newsgroup,
use the UNIX program rn,
the XWindows version, xrn,
any of a variety of Macintosh newsreaders, or a worldwide web browser. The caenhelp facility has documentation.
There will be six homework
assignments (problem sets) in this course, put on the web (at http://ai.eecs.umich.edu/people/durfee/courses/492winter01/assignments)
and due at approximately twoweek intervals. Some of these will involve some
programming or other software work. In addition to programming, assignments
might also involve mathematical analysis, information gathering (from WWW and
elsewhere), and critical thinking. Do not leave problem sets (especially
programming tasks) to the last minute, as network servers sometimes go down,
and working through the problems usually takes longer than you think (even when
you take this fact into account). We give you two weeks for a reason!
All assignments are to be handed in no
later than the start of class (9:10 am) on the due date. If you hand your
assignment in after the class has begun, it is late, and will be subject to the
late policy. Be sure to get your
homework in on time. Unless
otherwise directed in the assignment, you should turn in a hardcopy of the
assignment. You should not put
your assignment in an unsecured dropoff point (such as in an instructor’s mail
slot in the EECS or ATL building); instead, you should hand your assignment in
directly to an instructor (before lecture, at office hours or discussion
section, etc.), or in a secure place (e.g., under the professor’s locked office
door). (Warning: Access to the ATL
building is limited between 5pm and 8am.)
On the morning the assignment is due, it must be handed in to the
instructors.
We expect strict adherence to the
Engineering Honor Code in all assignments and exams. All problem sets (home
work assignments) are to be completed on your own. You are encouraged to
discuss ideas and techniques broadly with other class members, but all written
work, whether in scrap or final form, are to be generated by you working alone
unless otherwise expressly stated in the homework assignment. You are not
allowed to sit together and work out the details of the problems with anyone.
You are not allowed to discuss the problem set with previous class members, nor
anyone else who has significant knowledge of the details of the problem set.
Nor should you compare your written solutions, whether in scrap paper form or your
final work product, to other students (and vice versa), material on the web,
assignments from previous years, and so forth. You are also not allowed to
possess, look at, use, or in anyway derive advantage from the existence of
solutions prepared in prior years, whether these solutions were former
students' work product or copies of solutions that had been made available by
instructors. Violation of this policy is grounds to initiate an action that
would be filed with the Dean's office and would come before the College of
Engineering's Honor Council. If you find any ambiguity about this policy, it is
your responsibility to contact the course staff.
To be considered on time,
assignments must be turned in by the start of class (9:10am) on the due date.
You are welcome to submit your answers earlier, certainly. Late assignments
will be assessed a penalty of 10% of the assignment's available points per day
or fraction thereof, up to a maximum of three days after the due date. If
necessary due to the course schedule, some assignments might not have the
possibility of three late days; in such cases, the assignment will expressly
say so. Since we generally provide this routine slack period, we will not
grant extensions except under the most extraordinary of circumstances.
Problem sets 60% (10% each); Exams
40% (midterms 12% each, final 16%).
(These percentages are subject to slight adjustments at the discretion
of instructor.)
If you have a problem with the
grading on a particular assignment, write a brief (oneparagraph) description
of the problem, and hand it with the assignment to the GSI in section for a
regrade. Regrade requests must be submitted within one week of when the graded
assignment is made available to the student. Later regrade requests will
not be accepted.
There will be three
examinationstwo midterms and a final.
The midterms will be held during class period; on those class days the expectation
is that the exam will begin at 9AM sharp.
Please ensure that you do not miss the exams, and that you arrive
promptly for them.
Midterm#1: Thursday, 15 February,
9:0010:30AM, 12% of grade.
Midterm#2: Tuesday, 27 March,
9:0010:30AM, 12% of grade.
Final: Monday, 23 April,
1:303:30, 16% of grade.
(Subject to
midcourse adjustments). Readings are from Russell&Norvig. PS = Problem
Set.
Warning: The week of February 13,15
will be VERY busy. Be ready!
Date 
Topic 
Readings 
Homework 
4 January 
Introduction and
Overview 
Chap 1 
PS1 out 
9 January 
Intelligent Agents
and Problem Solving 
Chap 2 

11 January 
Problem Solving 
Chap 3.03.3 

16 January 
Blind Search 
Chap 3.4+ 
PS1 due, PS2 out 
18 January 
Informed Search 
Chap 4.04.2 

23 January 
Memory Bounded
Search 
Chap 4.3+ 

25 January 
Reasoning in
Propositional Logic 
Chap 6.06.4 

30 January 
FirstOrder Logic 
Chap 7.07.3 
PS2 due, PS3 out 
1 February 
Logical Agents 
Chap 6.5+, 7.4+ 

6 February 
FirstOrder
Inference 
Chap 9 

8 February 
Planning 
Chap 11.011.5 

13 February 
Partial Order
Planning 
Chap 11.6+ 
PS3 due, PS4 out 
15 February 
Midterm 1 

Midterm1 
20 February 
Other Planning
Topics 
Chap 12 

22 February 
Plan Execution 
Chap 13 

27 February 
Spring Break! 


1 March 
Spring Break! 


6 March 
Uncertainty and
Probability 
Chap 14 
PS4 due, PS5 out 
8 March 
Bayesian Networks 
Chap 15.015.3 

13 March 
Probabilistic
Reasoning 
Chap 15.4+ 

15 March 
Decision Making 
Chap 16 

20 March 
Machine Learning:
Decision Trees 
Chap 18.018.4 
PS5 due 
22 March 
Machine Learning:
Hypothesis Space 
Chap 18.5+ 

27 March 
Midterm 2 

Midterm2, PS6 out 
29 March 
Learning in Neural
Networks 
Chap 19 

3 April 
Learning Reprise 


5 April 
Agent Modeling 
Chap 22.122.2 

10 April 
Game Tree Search 
Chap 5 

12 April 
Language and
Communication 
Chap 22 

17 April 
Perspective 
Chap 27 
PS6 due 
23 April 
Final Exam 1:303:30 PM 

