CIS 830: Advanced Topics in Artificial Intelligence
Spring 2000
Hours: 3 hours (extended course project option for 4
credit hours: 3 of CIS 830, 1 of CIS 798)
Prerequisite: First
course in artificial intelligence (CIS 730 or equivalent coursework in theory
and design of intelligent systems) or instructor permission
Textbook: Artificial Intelligence: A Modern
Approach, S. J. Russell and P. Norvig.
Prentice-Hall, 1995. ISBN:
013108052
Venue: Monday, Wednesday, Friday 10:30-11:20 Room 236 Nichols Hall
Instructor: William H. Hsu, Department of Computing and Information Sciences
Office: 213 Nichols Hall URL:
http://www.cis.ksu.edu/~bhsu E-mail: bhsu@cis.ksu.edu
Office phone: (785) 532-6350 Home
phone: (785) 539-7180
Office hours:
before class (10-10:30am); 2-2:30pm, Wednesday; 1:30-2:30pm, Friday; by
appointment
Class web page: http://ringil.cis.ksu.edu/Courses/Spring-2000/CIS830/
This course
provides intermediate background in intelligent systems for graduate and
advanced undergraduate students. Five
topics will be discussed: analytical and multistrategy learning (hybrid
analytical, knowledge-driven, and inductive learning), pattern recognition
using artificial neural networks, reasoning under uncertainty for decision
support, data mining (DM) and knowledge discovery in databases (KDD), and
genetic algorithms and evolutionary systems for KDD. Applications to practical design and development of intelligent
systems will be emphasized, leading to class projects on current research
topics.
Homework: 4
programming and written assignments (30%)
Papers: 15 (out
of 20) written (1-page) reviews of research papers (15%); class presentations
(15%)
Examinations: 1
take-home midterm (15%)
Computer language(s): C/C++,
Java, or student choice (upon instructor approval)
Project: term programming project for all
students (25%); additional term paper or project extension (4 credit hour)
option for graduate students and advanced undergraduates
·
Neural Networks for Pattern Recognition, C.
M. Bishop. Oxford University Press, 1995. ISBN: 0198538499
·
Genetic Algorithms in Search, Optimization, and Machine Learning, D. E. Goldberg. Addison-Wesley, 1989. ISBN: 0201157675
·
Machine Learning, T. M. Mitchell. McGraw-Hill,
1997. ISBN: 0070428077
·
Advances in Knowledge Discovery and Data Mining, U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, eds. MIT Press,
1996. ISBN: 0262560976
·
Readings in Machine Learning, J. W.
Shavlik and T. G. Dietterich, eds. Morgan-Kaufmann, 1990. ISBN: 1558601430
·
Readings in Computer Inference and Knowledge Acquisition, B. G. Buchanan and D. C. Wilkins, eds. Morgan Kaufmann, 1993. ISBN:
1558601635
·
Probabilistic Reasoning in Intelligent Systems, J. Pearl. Morgan-Kaufmann,
1988. ISBN: 0934613737
·
Genetic Programming, J. Koza. MIT Press, 1992.
ISBN: 0262111705
·
Pattern Classification and Scene Analysis, R. O. Duda and P. E. Hart. John Wiley and Sons, 1973. ISBN: 0471223611
|
Lecture |
Date |
Topic |
(Primary) Source |
|
0 |
January 14 |
Administrivia; overview of topics |
RN Chapters 1, 14, 18 |
|
1 |
January 19 |
Analytical learning: overview |
RN 21 |
|
2 |
January 21 |
Analytical learning: presentation |
Thrun/Mitchell |
|
3 |
January 24 |
Analytical learning: discussion |
Thrun/Mitchell; RN 21.2 |
|
4 |
January 26 |
Analytical learning: presentation |
Brodie/Dejong; TMM 11 |
|
5 |
January 28 |
Analytical learning: discussion |
Brodie/Dejong |
|
6 |
January 31 |
Analytical learning: presentation |
Chown/TGD; TMM 12 |
|
7 |
February 2 |
Analytical learning: discussion |
Chown/Dietterich |
|
8 |
February 4 |
Analytical learning: presentation |
Towell et al (TSN) |
|
9 |
February 7 |
Analytical learning: conclusions |
TSN; RN 21; TMM 12 |
|
10 |
February 9 |
Artificial neural networks: overview |
RN 19; TMM 4 |
|
11 |
February 11 |
ANNs: presentation |
Maclin/Shavlik |
|
12 |
February 14 |
ANNs: discussion |
Maclin/Shavlik |
|
13 |
February 16 |
ANNs: presentation |
Sutton; RN 16 |
|
14 |
February 18 |
ANNs: discussion |
Sutton |
|
15 |
February 21 |
ANNs:
presentation |
Hinton et al |
|
16 |
February 23 |
ANNs: discussion |
Hinton et al |
|
17 |
February 25 |
ANNs: presentation |
Jordan/Jacobs |
|
18 |
February 28 |
ANNs:
conclusions |
Jordan/Jacobs;
RN 16 |
|
19 |
March 1 |
Uncertainty: overview |
RN 15; TMM 6 |
|
20 |
March 3 |
Uncertainty: presentation |
Cheeseman |
|
21 |
March 6 |
Uncertainty: discussion;
midterm review |
Cheeseman |
|
22 |
March 8 |
Uncertainty: presentation; midterm |
Friedman et al |
|
23 |
March 10 |
Uncertainty: discussion |
Friedman
et al |
|
24 |
March 13 |
Uncertainty: presentation |
Horvitz et
al; RN 17 |
|
25 |
March 15 |
Uncertainty: discussion; midterm due |
Horvitz et
al |
|
26 |
March 17 |
Uncertainty: presentation |
Darwiche/Pearl |
|
27 |
March 27 |
Uncertainty: conclusions, DM:
overview |
Darwiche/Pearl |
|
28 |
March 29 |
DM/KDD/decision support:
presentation |
Fayyad/Haussler/Stolorz |
|
29 |
March 31 |
DM/KDD/decision support: discussion |
Fayyad/Haussler/Stolorz |
|
30 |
April 3 |
DM/KDD/decision support:
presentation |
John/Kohavi/Pfleger |
|
31 |
April 5 |
DM/KDD/decision support: discussion |
John/Kohavi/Pfleger |
|
32 |
April 7 |
DM/KDD/decision support:
presentation |
Koller/Sahami |
|
33 |
April 10 |
DM/KDD/decision support: discussion |
Koller/Sahami |
|
34 |
April 12 |
DM/KDD/decision support:
presentation |
Craven et
al |
|
35 |
April 14 |
DM/KDD/decision support: conclusions |
Craven et
al |
|
36 |
April 17 |
Genetic algorithms: overview |
TMM 9 |
|
37 |
April 19 |
GAs: presentation |
Wilson |
|
38 |
April 21 |
GAs: discussion |
Wilson |
|
39 |
April 24 |
GAs: presentation |
Aler/Borrajo/Isasi; Koza |
|
40 |
April 26 |
GAs: discussion |
Aler/Borrajo/Isasi |
|
41 |
April 28 |
GAs: presentation |
Raymer et
al |
|
42 |
May 1 |
GAs: discussion |
Raymer et
al |
|
43 |
May 3 |
GAs: presentation |
Horn et
al |
|
44 |
May 5 |
Conclusions; all
projects due |
Horn et al; TMM 9 |
|
45 |
May 8 |
NO FINAL EXAM |
RN Sections V-VII |
RN: Artificial Intelligence: A Modern Approach, S. J. Russell and P.
Norvig
TMM: Machine Learning, T. M. Mitchell
Lightly-shaded entries denote
the due date of a paper review.
Heavily-shaded entries denote the due date of a paper review and a written or programming assignment.
·
Official class page: http://ringil.cis.ksu.edu/Courses/Spring-2000/CIS830
·
Instructor’s home page: http://www.cis.ksu.edu/~bhsu
Note: It is the
student’s responsibility to be aware of class announcements and materials
posted on the official class page, so please check it frequently.
Required
readings, along with reference manuals and tutorials for software used in the
course, will be available for purchase (in 2 packets) from the Engineering Copy
Center in 14 Seaton Hall.
·
URL: http://ringil.cis.ksu.edu/Courses/Spring-2000/CIS830/Board
·
Primary purpose: for class discussions (among students and
with instructor)
Note: Postings
on the web board will tend to get a more rapid response from the instructor
than e-mail, besides which, they are sometimes of benefit to fellow students.
·
Tentative title: CIS830WHH-L@cis.ksu.edu
·
Optional (no course-critical announcements)
·
For research announcements, related job opportunities, etc.
·
Sign up if interested
Homework assignments will be given out 2 to 3 weeks
apart, for a total of 4. Your lowest
score will be dropped (see below). One
of these homeworks will be programming-based; one will require you to run (and
possibly modify) an existing library or KDD package using a specification or
sample data, and analyze the results; and two will be written.
Type (do not hand-write) homeworks; handwritten
solutions are worth 0.8 credit.
For programming assignments and the course project,
you are permitted to use your choice of a high-level programming language (C++
and Java are strongly preferred; consult the instructor if you intend to use
any other programming language). You
must, however, use a development environment that is available to the CIS
department. Consult the class web page
for approved compilers.
The course project proposals will be due on Monday,
February 14, 2000 and will count for 20% of the project grade.
For graduate students and advanced undergraduates interested in working on a class project, you may elect an additional 1 hour of credit as a section of CIS 798 (Special Topics in Computer Science) and either turn in a term paper or work on an extension of the course project or a small-scale independent study project. You may sign up for this option any time before February 14, 2000 (talk to me during office hours or send e-mail). Suggested project topics and guidelines will be posted on the course web page. Examples include: improving a known supervised learning algorithm; developing an algorithm for combining classifiers; a KDD application of DBMS or OLAP, in support of an MSE concentration in database systems; a project on analysis of time series or document databases (e.g., text or source code); an in-depth comparison of two KDD techniques studied in the course; or improving an existing model or analyzing it formally.
Cheating consists of misrepresenting another’s work
or knowledge as your own. It includes not
only copying of test answers, but plagiarism of another person’s written
material. While you are encouraged to
discuss class material, homework problems, and projects with your classmates,
the work you turn in must be entirely your own. For homework assignments, this means that if you work together
with a fellow student, you should still produce the final, written material
from your own notes and individual work, rather than from common notes that you
produced together. You should follow
similar guidelines for programming assignments and individual projects; while
reuse of previously developed source codes may be permitted in these cases
(provided you acknowledge the authors appropriately), you must not use directly
use code developed by fellow students.
Please consult the University honor code (http://www.ksu.edu/honor) for further
guidelines on ethical conduct, and understand the regulations and penalties for
violating them.
The codes that you are permitted to use on certain assignments may be limited, beyond the specifications of plagiarism standards. When in doubt about whether you may use a particular program on a written or programming assignment, consult the instructor first. My objective is to help you learn as much as possible from the assignments; sometimes this means that I want you to use existing code and sometimes I will prefer for you to develop it yourself, to better understand the techniques.
Credit
for the course will be distributed as follows:
|
Component |
Quantity
|
Low Scores Dropped
|
Points Each (Out of 1000)
|
Value
|
|
Homework (Written/Programming Assignments) |
4 |
1 |
100 |
30% |
|
Paper Reviews and Commentaries |
20 |
5 |
10 |
15% |
|
Presentation |
1 |
0 |
150 |
15% |
|
Midterm Exam (Take-Home, Open-Book) |
1 |
0 |
150 |
15% |
|
Course Project (Due May 5, 2000) |
1 |
0 |
250 |
25% |
Homework
and exams may contain extra credit
problems.
Late policy: Homeworks are due at 5:00pm
on Fridays; you may request an extension to the following Monday if you need
one by the due date (but I recommend
you do not take this option). 10%
credit will be deducted for each day the assignment is late past 5:00pm that
Monday. There will be no
additional extensions!
Letter grades will be assigned based on the
distribution of raw scores (“curved”). Undergraduate and graduate students will
be graded on the same curve. Acquiring
85% of the possible points, however, guarantees an A; 70%, a B; 55%, a C. Actual scales may be more generous than this
if called for, but are not expected to be.
If you elect to take an additional CIS 798 project option (for 1 hour of credit), your grade for CIS 830 will still be assigned based only on the above components. The additional project component will be graded separately (as CIS 798) and weighted proportionately.
1. Assigned Monday, January 31, 2000, due Friday, February 25, 2000
2. Assigned Monday, February 28, 2000, due Friday, March 17, 2000
3.
Assigned Monday, March 27, 2000, due
Friday, April 7, 2000
4.
Assigned Monday, April 10, 2000, due
Friday, April 28, 2000
An important part of learning
about intelligent systems and knowledge discovery in databases, whether for
research or development applications, is understanding the state of the field
and the repercussions of important results.
The readings in this course are designed to give you not only a set of
tutorials and references for machine learning tools and techniques, but to
demonstrate the subject as a unified whole, and to encourage you to think more
deeply about the practical and theoretical issues.
Toward this end, I have selected 4
papers out of those in your (2) course notes packets. The first 2 of these are in the first packet and the last 2 are
in the second. Before you come to
lecture on the dates indicated on the class calendar, you should submit (by
e-mail to the instructor) a short
review of, and commentary on, the assigned paper. This commentary need be
no longer than 2 pages (though you can go up to 3 pages if you feel you have
something meaningful to add).
This review is an important part
of the course, because it can:
·
help you to review and rehearse material from lecture
·
bring to light questions that you may have about the
material
·
improve your ability to articulate what you have learned
·
help guide the lecture/discussion
·
help you to think about working on projects (implementations
or research) in this field
Here are some
guidelines on writing the reviews:
1. Try to be
brief and concise.
2. Concentrate
on pointing out the paper’s main strengths and flaws, both in content
(key points, accuracy, impact/implications, deficiencies) and in presentation (organization,
clarity/density, interest). Try not
to merely summarize the paper.
3. Some
questions to address (typically a couple in each paper):
·
Is the paper of sufficiently broad interest? What do you think its intended audience is? To
what audience do you think the paper is significant?
·
What makes the paper significant or insignificant?
·
How could the presentation be improved to better point out
important implications?
·
Is the paper technically sound? How so, or in what areas is it not entirely sound?
·
What novel ideas can we pick up (about the topics covered in
lecture) from the paper?
·
Comment on how the paper (or the topic) affects your own
work. How is it relevant (or
irrelevant) to you?
4. How might
the research be improved in light of how the field has progressed since it was
published? Some of these papers were
catalysts for research in their areas, so it is sometimes infeasible to
second-guess their authors; but comment on what could be done better today.
Paper reviews are late (worth 0 credit) after midnight of
the day of the lecture when they are due (i.e., you must submit them before
12:00am Tuesday, Thursday, or Saturday).
Do not plagiarize. It is relatively easy to detect plagiarism of material from the paper itself, related references, and paper reviews of classmates! Again, refer to http://www.ksu.edu/honor for regulations and further guidelines on academic honesty.