Machine learning
4 points * Two 1-hour lectures per week * Second semester * Clayton * Prerequisites: As for CSC3010; additional prerequisite CSC3091 * Corequisites: As for CSC3030
This subject examines the main contending approaches taken to the computational modelling of learning. Topics include: production systems as cognitive models and their evolution into expert systems; version space approaches to concept formation; explanation-based learning; `scientific discovery' (data-driven) methods of learning empirical laws. It has been argued that symbolic AI cannot cope with some varieties of learning tasks; therefore, the subject takes up various quantitative methods of learning as well, especially information-theoretic methods of classification and Bayesian learning techniques. Also covered are: alternative non-symbolic learning models offered by evolutionary programming (genetic algorithms) and connectionism (neural networks), and the alleged incompatibilities between these different models.
Assessment
Examination (2 hours): 50% * Programming project: 50%
Prescribed texts
Thornton C J Techniques in computational learning Chapman and Hall, 1992
Recommended texts
Carbonell J G (ed.) Machine learning: Paradigms and methods, MIT, 1989
Shavlik J W and Dietterich T G (eds) Readings in machine learning Morgan Kaufmann, 1990
Shrager J and Langley P (eds) Computational models of scientific discovery and theory formation Morgan Kaufmann, 1990