MACHINE LEARNING
			       CSC3200

			Instructor: K. Korb

4 points.  Two 1-hour lectures per week.  Second semester, Clayton.
Prerequisites: CSC3091.

Objectives.  On completion of the subject students will be able to:


appreciate the role of machine learning in AI theory and
applications;
understand the main current techniques used to implement
machine learning;
design and implement a machine learning program.


Synopsis 

This course examines the main approaches taken to the computational
modeling of cognition, emphasizing machine learning.  We begin with an
examination of production systems as cognitive models and their
evolution into expert systems.  A study of their limitations leads to
the major contending paradigms of machine learning.  Among symbolic
approaches to learning we consider: version space methods;
explanation-based learning; `scientific discovery' (data-driven)
methods.  Symbolic AI has difficulties dealing with some varieties of
learning tasks; therefore, we also take up: information-theoretic
classification, Bayesian learning, evolutionary algorithms, and
connectionism.  We will also consider the alleged incompatibilities
between these different models.


Assessment. Examination: 2 hours (50%); Project (50%).

Prescribed texts

C.J. Thornton Techniques in Computational Learning.  Chapman and Hall, 1992.

Recommended texts

J.W. Shavlik and T.G. Dietterich (eds.) Readings in Machine Learning.  
   Morgan Kaufmann, 1990.

J. Shrager and P. Langley (eds.) Computational Models of Scientific 
   Discovery and Theory Formation.  Morgan Kaufmann, 1990.

J.G. Carbonell (ed.) Machine Learning: Paradigms and Methods.  MIT, 1989.

R. Wilensky COMMON Lispcraft, 3rd ed. Norton, 1986.

Patrick Winston & Berthold Horn Lisp, 3rd ed. Addison-Wesley, 1989.