Textbook:  Artificial Intelligence, A Modern Approach: Second Edition.  Stuart Russell & Peter Norvig.  Prentice Hall, 2003.  Please note that you must have the Second Edition.

CSE 326 required, CSE 341 recommended.  Not open for credit to students have have completed CSE 415.

Knowledge Post-Conditions

This course will enable you to:

• Represent real-world constraint satisfaction problems in terms of state-space search, and solve them using a variety of different search algorithms.
• Represent and solve deductive reasoning problems using propositional logic.
• Represent and solve probabilistic reasoning problems using Bayesian networks.
• Use machine learning algorithms, including neural networks and decision trees, to solve classification problems such as arise in medicine and science.
• Understand how perception, learning, and reasoning interact in an autonomous AI system.

Course Work

A set of assignment will be given out (almost) every Monday.  The due date for each task will be specified.  It is important that you finish each task by its due date, because doing the task prepares you for the work we do in the classroom that day.  There will be 5 different kinds of tasks:

• Readings:  Read the assigned pages from the texbook and write notes that summarize the material.  The notes may be as short or long as you like, typically one side of a sheet of paper.    The readings and dates due are all specified on the Calendar page.  On the indicated date turn in a short summary of the readings at the start of class.  These will be checked off as having been done, and returned to you.  For the final exam you may bring these notes plus any others you take during or after class.
• Written Exercise:  A paper and pencil type exercise.  Answers may be handwritten or typed.  These will be graded.
• Programming Problem:  These tasks will involve implementing a program and testing it on a given set of data.  Programming tasks should be done in teams of two.  I will prepare the assignments with a Java implementation in mind, but you are free to use a different programming language (for example, C++, LISP, ML, Prolog, or Python).  Data sets and support code (if any) will be provided on the instruction Unix servers.  You may transfer these materials to a Windows machine and work in that environment if you prefer.  Programming tasks will be graded on correctness, documentation and understandability of the code, and write-up of the experimental results.
• Application of Software Tools:  These tasks will involve learning to use a AI software tool (for example, the C4.5 decision tree learning system) and analyzing the result of using the tool on a set of test data.  These tasks will be graded on correctness and the write-up.

Grading

70% of your grade will be based on the assignments and 30% on the midterm and final exam.  See the Grade Breakdown page for the formula used to calculate the final grades.

The midterm exam will be in class on Nov 5.  The final exam is 8:30 am - 12:20 pm on Friday, Dec 12.

Staying in Touch

The course mailing list is <cse473@cs.washington.edu>, and will be enabled by Oct 3rd.  Please use it as a resource to discuss issues about the course and the assignments with other students.  The TA and instructor will also be reading this list, so it is also appropriate for any question of a general nature.  You may also send email to Jeffrey or me directly with either general questions or questions about your own performance in the course.  The course home page is:

http://www.cs.washington.edu/education/courses/473/03au

I encourage people to come to office hours.  You do not have to wait until you have a problem with an assignment:  I am happy to talk about your progress in the undergraduate program, career and graduate school opportunities, or wild and crazy ideas about AI.  I will be out town at computer science research conferences Sept 30-Oct 3, Oct 20, and Nov 10-12.  Jeffrey and guest lecturers will fill in on these days.

Collaboration and Academic Honesty

In this class, you will be collaborating with other students to solve problems. Some solutions will be handed in and graded as a group activity and some will be individually written and evaluated. When doing individual assignments, you may still collaborate with other students. However, you must give credit to others who have given you some insight into solving a problem, just as professionals recognize others who have helped with their research or writing.