Objectives

To gain a working knowledge of the foundations of, and modern applications in, artificial intelligence, including agent design, heuristic search, knowledge representation, planning, logic, natural language processing and machine learning. Students will gain experience writing several AI applications in a variety of programming languages.

Prerequisites

Completion of courses in Data Structures, Algorithms, and Computation Theory, and two to three years of programming experience. Some exposure to functional or scripting languages is a plus. Dr. Dionisio's class on Computer Graphics will be helpful preparation for the robotics portion of the course. Non-computer science majors that lack these prerequisites will take a separate, independent track through the material and will write an extensive term paper about novel applications of AI in their field of study.

Readings

The textbook for the class is:

• Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach, 2nd edition, Prentice Hall, 2003.

Recommended Readings:

• Alan Turing, Computing machinery and intelligence, Mind  59:433–460, 1950.
• Douglas R. Hofstadter, Gödel, Escher, Bach: An Eternal Golden Braid, 20th anniversary edition, Basic Books, 1999.

Additional papers and readings will be assigned throughout the course (including my own course notes, practice problems, and sample code). If you have projects or papers to work on, you'll have to find some additional readings on your own. Use judgment when researching on the web; a fair amount of information is often wrong, and much of the so-called sample code is especially atrocious. Regardless, you must take the time for effective self-study that includes practicing the craft of programming.

Assignments and Grading

You'll have several homework sets containing in-depth theoretical questions and non-trivial programming problems, and quizzes and a final exam with less difficult material. There will be a small project to work on throughout the term; students are encouraged to build a little robot that can do something interesting. To help prepare you to meet industry expectations for college graduates, most assignments will take the form of open source software products. Unless otherwise specified, you are required to keep all work in your CVS repository and prepare all homework solutions with LaTeX document preparation system. Exams will cover material from lectures not previously assigned for homework: don't whine about this.

Generally, coursework may be done in groups of no more than two students; however, while only one solution set is turned in per group, both students are responsible for understanding all of its content and may be asked at any time for an oral explanation of any solution. Collaboration with other groups is fine but must be limited: you may share ideas and approaches but nothing resembling a solution (not even pseudocode). You must also acknowledge any help received. Academic dishonesty may result in expulsion; be certain your work meets the standards set forth in the LMU Honor Code.

Your final grade will be weighted as follows:

Letter grades are figured according to the usual scale: 90% or more of the total points gets you an A, 80% a B, 70% a C, and so on. These are minimal requirements; for example, if you get 82 points you are guaranteed a B- or better, though you might still get an A since 82 may be the top score.

Homework is due at the beginning of class; late assignments are docked 30% per class. Missing class just to get an assignment done on time will not be tolerated; the only good excuses for missing class are excellent surf conditions, family problems, sickness, and personal emergencies. Skipping class just puts your fellow students at an advantage: we often spend class time going over things that will be "on the exam".

Your programming style will play a huge part in determining your score on the programming assignments. I will not hesitate to assign D's or F's to working programs which are poorly structured, under-commented, have poor identifier names and abbreviations, contain inappropriate hard-coded values, or are not easily maintainable. Appearance of the grading policy in this syllabus constitutes fair warning of the consequences of poorly written code.

Topics and Tentative Schedule

• INTRODUCTION TO AI: Definitions; Concerns, areas of study, brief history, and application areas. (Ch. 1)
• AGENTS: Agent structure, behavior, and implementation; Performance, omniscience, and rationality; Task environments; Agent design. (Ch. 2)
• PROBLEM SOLVING: Problem solving agents; The notion of search; Uninformed search strategies: BFS, DFS, Iterative deepening, Bidirectional search; Heuristic search strategies: Best-first, A*, Hill-climbing, Beam search; Simulated annealing; Constraint satisfaction; Adversarial search: Minimax, α-β, Games, etc. (Ch. 3, 4, 5, 6)
• KNOWLEDGE AND INFERENCE: Knowledge-based agents; Logic: Propositional logic, First-order logic: quantifiers, unification, forward and backward chaining, resolution; Higher-order logic; Knowledge representation and ontology: situation, time and event calculi, belief theory. (Ch. 7, 8.1–8.3, 9, 10.1–10.4)
• PLANNING: Problems and decomposability; Planning languages; Planning and heuristics in state-space search; Partial order planning. (Ch. 11.1-11.3)
• UNCERTAINTY: Uncertain knowledge, uncertainty and rationality, Probability basics: propositions, events, axioms, inference, independence; Bayes' Rule; Bayesian Networks: structure, exact inference and approximate inference; Temporal uncertainty: Hidden Markov models, Kalman filters, and Dynamic Bayesian networks; Speech Recognition. (Ch. 13, 14.1–14.5, 15)
• UTILITY AND DECISION MAKING: Utility Theory, utility functions; Decision networks; The value of information; Sequential decision problems: value and policy iteration. (Ch. 16, 17.1–17.6)
• LEARNING: Overview of learning; Inductive learning; Decision Tree learning; Statistical learning; Neural networks; Reinforcement Learning: passive and active approaches, generalization. (Ch. 18.1–18.3, 21.1–20.2; 20.5, 21.1–21.4)
• COMMUNICATION: Multi-agent environments; Basic language theory; Natural Language: syntax analysis, semantics, ambiguity, natural language understanding (Ch. 22)
• ROBOTICS: Perception and action; Image processing: basic operations, edge detection, recognizing gradients and shading, 3-D object reconstruction from 2-D images; Robotic hardware; robotic planning; robotic motion; robot software. (Ch. 24, 25)
• PHILOSOPHY, ETHICS, AND THE FUTURE OF AI: Strong vs. Weak AI; Three philosophers who don't like AI (more or less); Ethics and risks; Current research; Future directions; Your project presentations! (Ch. 26, 27)
• FINAL EXAM: 11:00 a.m. on Tuesday, May 2, 2006

University Links

All students will want to acquaint themselves with the useful information found in the following sources:

• LMU Academic Requirements and Policies (Pay particular attention to the sections on academic dishonesty)
• LMU Disability Support Services Office
• LMU Student Codes and Policies
• LMU Learning Resource Center