Syllabus
CSCI 3202
Artificial Intelligence
Tu, Th 12:30 - 13:45
ECCR 150
Instructor
Teaching Assistant
Course Objectives
This course surveys techniques involved
in creating computer systems that engage in intelligent human-like
behavior. Although Artificial Intelligence (AI) has received a
lot of play in the media, the true state of AI research is nowhere near
the vision depicted in the movie of the same name, or in the Matrix and Terminator series. The goal
of this course is to present a realistic view of the state of AI
techniques and research, and for students to develop small systems that
exhibit learning and intelligent behavior.
Although AI as a field has been around since the 1960's, in the past
decade it has undergone a radical transformation. The traditional
approach to AI was based on symbol
processing: the mind was conceived of as a symbol processing engine that
creates and manipulates symbol strings. The modern approach to AI
treats the mind a statistical
and probabilistic computing engine. CSCI 3202 will
emphasize the modern approach.
Prerequisites
CSCI 3104 and CSCI 3155 are the
official prerequisites. However, it will also be useful for you
to have had linear algebra and some background in probability and
statistics.
Course Materials
Text
The course text is
Artificial
Intelligence: A Modern Approach (Second Edition) by Russell
& Norvig. This text is difficult but thorough and
state-of-the-art. We will focus on the second half of the text,
which covers current trends in AI. Some resources related to the
text, including code, are available
here.
Computers
You may do homework assignments using
any programming language that you are comfortable with, and you may use
your own machine, the CSEL facilities, or the ITS facilities on
campus. Please see the TA if you need help gaining access
to computing facilities.
Course Schedule
Class
Date
|
Lecture
Topic
|
Reading |
Homework
|
Aug
24
|
Introduction
|
|
Homework
1 assigned
|
Aug
26
|
Examples
of
AI systems
|
|
Homework
1 due
|
Aug
31
|
History
of AI
|
Ch.
1
|
|
Sep
2
|
Intelligent
agents
|
Ch.
2, Section 7.2
|
Homework
2 assigned
|
Sep
7
|
Representing
uncertainty I
|
Ch.
13 (don't worry about logic rules on p. 463); Section A.3
|
|
Sep
9
|
Representing
uncertainty II
|
Homework
2 due; Homework 3 assigned
|
Sep
14
|
Reasoning
under uncertainty I
|
Ch.
14
|
|
Sep
16
|
Reasoning
under uncertainty II
|
Homework
3 due
|
Sep 21
|
Reasoning
under uncertainty III
|
Homework
4 assigned
|
Sep
23
|
Simple
decision making
|
Ch.
16
|
|
Sep
28
|
Sequential
decision making
|
Sections
17.1-3
|
|
Sep
30
|
No
class -- Fall break
|
|
|
Oct
5
|
Reinforcement
learning I
|
Ch.
21
|
|
Oct
7
|
Reinforcement
learning II
|
|
Homework
4 due;
Homework
5
assigned
|
Oct
12
|
Supervised
learning I
|
Sections
18.1-3
|
|
Oct
14 |
Supervised
learning II
|
Section 20.4
|
|
Oct
19 |
Supervised
learning III
|
Sections
20.1-2
|
Homework
5 due; Homework 6 assigned
|
Oct
21
|
Neural
networks I
|
Section
20.5; Section A.2
|
|
Oct
26
|
Neural
networks II
|
|
|
Oct
28
|
Neural
networks III
|
|
Homework
6 due; Homework 7 assigned
|
Nov
2
|
Amplifying
human capabilities with AI (Gerhard
Fischer)
|
to be announced |
|
Nov
4
|
Application: Data mining
|
Mozer et
al. (2000) |
|
Nov
9
|
Application:
Adaptive control
|
Mozer
(2004) |
|
Nov
11 |
Ensemble
methods
|
Section
18.4
|
|
Nov
16
|
Speech
recognition
|
Chapter
15
|
Homework
7 due;
Homework 8 assigned |
Nov
18
|
Exploiting
chaos (Liz Bradley)
|
web link
|
|
Nov
23
|
Cognitive modeling
|
Mozer
et al. (2004)
|
|
Nov
25
|
Thanksgiving
-- No Class
|
|
|
Nov
30
|
Game
AI: Collaborating agents in game design (Alex
Reppenning)
|
to be announced
|
|
Dec
2
|
Vision
(Jane Muilligan)
|
Chapter
24
|
Homework
8 due
|
Dec
7
|
Robotics
(Greg Grudic)
|
Chapter
25
|
|
Dec
9
|
Philosophical issues; Directions of
AI research
|
Chapters
26, 27
|
|
Course Requirements
Reading
Reading assignments from the Russell
& Norvig text are listed in the course schedule above. I
suggest that you attend lecture and then attempt to read the
corresponding material in the text, because the more mathematical
sections of the text are challenging, and I'll present the highlights
of the text which should help in understanding the details when you
read the text.
Lectures
Class attendance is mandatory. If
you find that you are not getting a lot out of lectures--either because
the material is too basic or too complex--it is your job to let me
know, and I'll adjust the lectures.
Homework Assignments
Although students may understand AI
techniques in principle by reading the book and attending lectures,
there is no substitute for the sort of understanding one obtains by
actually implementing a technique and trying it out on a problem, even
if the problem is small relative to the sort of problems we wish AI
system to tackle. To this end, the course will entail 8 homework
assignments. For the assignments inovlving programming, you may use
whatever language you are most comfortable with -- Java, C++, python,
matlab, etc.
Because the course schedule is tentative, homework due dates may
change. However, when homework is assigned, the due date will be
indicated, and it is
not
flexible. Late assignments will not be accepted.
Assignments must be turned in at the start of class on the assignment
due date. If you anticipate any difficulty turning in your
homework on time, let your professor know
at the time when it is assigned.
Tentatively, the homeworks
will be weighted as follows:
Homework
#
|
Topic
|
Weighting
|
1
|
Research
existing AI systems
|
5%
|
2
|
Wumpus
world agent
|
10%
|
3
|
Probability
theory
|
10%
|
4
|
Probabilistic
inference
|
10%
|
5
|
Reinforcement
learning
|
15%
|
6
|
Simple
classifiers
|
10%
|
7
|
Neural
networks
|
15%
|
8
|
Final
project
|
25%
|
Exams
We will have a final exam, at the time
scheduled by the university, which is Saturday December
11, 7:30 p.m. - 10 p.m. The exam will be cumulative, and will
focus on evaluating your understanding of material that was presented
in the course but not central to the homework assignments (e.g., guest
lectures).
Semester Grade
Semester grades will be based 80% on
the homework assignments, 20% on the final.
Academic Honesty
In accordance with the University
Honor Code,
you may not give nor receive unauthorized assistance on the
homework. "Unauthorized assistance" includes help on designing
solutions to homework problems or sharing of code. You are
encouraged to discuss course material with fellow students, but you are
not to use any code anyone else has written as your own. You
should limit your discussion of homework assignments to understanding
the requirements of the assignment. If you have questions about
how to approach the problem, or get stuck working on your solution, you
should talk to the professor or TA.