Derek Bridge

This elective module was taught three times between 1997 and 2000 to final-year undergraduates on the main B.Sc. computing degrees in UCC's Computer Science Department.

• Lectures: 24 * 1 hr lectures
• Problem classes: 11 * 1 hr tutorials in the Spring term
• Private study: 52 hrs (including revision)

There are no formal prerequisite courses but students need to be comfortable with an amount of theoretical Computer Science: propositional logic, the rudiments of first-order logic, and algorithm and problem complexity.

• 1½ hour written exam (worth 80% of the marks) in the Summer term;
• A research paper (worth 20% of the marks) to be written during the Spring term.

We look at the techniques that are used for building knowledge-based systems, including state space search, the use of various logics for state representation, the use of automated logical inference of various kinds (primarily for 'within-state' reasoning), domain-independent A.I. planning methods (for 'between-state' reasoning), and simple machine learning.

In 1997-98, we then looked at how to build knowledge-based system for tasks that require expertise when done by humans, concentrating on analysis tasks of the diagnosis or fault-finding variety. We saw how the techniques learnt in the first part of the course are applied.

In 1998-99, we looked instead at machine learning. We considered different types of machine learning and saw three case studies. We did this again in 1999-2000.

• Students should learn a raft of techniques that can be deployed when building a knowledge-based system.
• Students should gain some understanding of the issues that are raised when applying these techniques.

• (L1) What is a KBS?
• (L2-5) State space search.
• (L6-16) Logical representation and reasoning.
• (L16-17) Semantic nets and frames as logics.
• (L17-21) AI planning.
• (L22-24) Machine learning.

• Copies of overheads (gzipped postscript) by topic are available below:
  • What are knowledge-based systems? (10k)
  • State space search (22k)
  • The syntax and semantics of first-order logic (96k)
  • Resolution refutation on clausal form logic (30k)
  • Practical machine reasoning (search, Horn clauses, semantic nets, etc.) (33k)
  • A.I. planning (27k)
  • Machine learning (20k)
• The assignment, 1999-2000
• Copies of exams (gzipped postscript):
  • Sample exam 1998 (44k)
  • Summer exam 1998 (39k)
  • Autumn exam 1998 (42k)
  • Summer exam 1999 (38k)

The Boole library has many AI/KBS books, and I shall be asking for more. Of the books in its collection, the following may be of use:

• Shapiro, S.C. and Eckroth, D. (eds.): Encyclopedia of artificial intelligence, Wiley
• Ginsberg, M.L.: Essentials of artificial intelligence Morgan Kaufmann Publishers, 1993
• Genesereth, M.R. and Nilsson,N.J.: Logical foundations of artificial intelligence Morgan Kaufmann, 1988
• Gabbay, D.M. et al.: Handbook of logic in artificial intelligence and logic programming, vols.1-4, Clarendon Press, 1993-95
• Rich, E.: Artificial intelligence, McGraw-Hill, 1983
• Winston, P.H.: Artificial intelligence (2nd edn.), Addison-Wesley, 1984
• Jackson, P.: Introduction to expert systems (2nd edn), Addison-Wesley, 1990
• Stefik, M.: Introduction to knowledge systems, Morgan Kaufmann Publishers, 1995

The following are top-class modern books:

• Russell, S.J. and Norvig, P.: Artificial Intelligence: a Modern Approach, Prentice-Hall, 1995
  Associated web pages
  A second edition is scheduled for November.
• Nilsson, N.J.: Artificial Intelligence: A New Synthesis, Morgan Kaufmann, 1998