2010-2011 Autumn Semester - 10 Credits
Summary
Machine learning is an important subfield of artificial intelligence involving automated discovery of concepts from examples, or, more generally, of patterns in data. It also provides the conceptual underpinning for the increasingly important practical field of data mining. This module will survey some of the principal approaches to automated concept learning. Topics will include version spaces, decision trees, Bayesian learning, instance-based learning, rule set learning, linear models and neural nets, association rule mining, clustering, combining multiple models, computational learning theory and the evaluation of hypotheses.
Aims
The aims of this module are:- to describe the main approaches to automated concept learning
- to develop students' skills in designing and building serious artificial intelligence programs
Objectives
By the end of this course the students should:- understand the basic data representations used in machine learning, and a number of fundamental learning algorithms;
- be able to use these representations and algorithms in constructing simple applications which learn to recognise patterns in data;
- be able to evaluate the suitability of different learning techniques for various kinds of applications.
Prerequisites and Corequisites
Advisable: knowledge of logical and rule-based approaches to
automated reasoning, from COM3290.
Advisable: knowledge of AI techniques, from COM1080.
Content
Concept learning and version spaces: [ 2 lectures ]
Decision trees: [ 2 lectures ]
Evaluating Hypotheses: [ 2 lectures ]
Bayesian learning and Bayesian belief nets: [ 3 lectures ]
Instance-based learning: [ 1 lectures ]
Rule set learning: [ 3 lectures ]
Computational Learning Theory: [ 2 lectures ]
Linear Models and Neural Nets: [2 lectures]
Association Ruled Mining and Clustering [1 lecture]
Combining Multible Models [1 lecture]
Structure and Teaching Method
-
There will be two formal hour hour lectures per week, to a total of 20
lectures.
-
A third hour per week may be used for exercises, question and answer sessions,
or lab work.
Resources Required
Weka Toolkit.
Assessment
-
Written examination (2 hours) at the end of the semester [70%].
- Two Assignments [30%]
How and When Students Will Receive Feedback
Students will receive written feedback on their assignments. Weekly tutorial sessions will discuss problem sheets assigned in advance and students will have the opportunity to present solutions and get oral feedback on these.
Recommended Texts
-
T.Mitchell , Machine Learning, WCB/McGraw Hill, 1997.
- I.H. Witten, E. Frank, Data Mining: Practical Machine Learning Tools and Techniques, 2nd Edition, Elsever, 2005.