Spring semester.

Unit(s) of Assessment	Weighting Towards Module Mark( %)
2 hour unseen examination	60%
Report on the practical application of machine learning techniques to a specified problem.  For a supplied data set, students must pre-process the data and systematically use it for training and testing of a neural network.  Various configurations of the system must then be used to obtain a good test performance on the data.   To address the following learning outcomes:   Apply machine learning techniques to solve complex problems.   Evaluate the use of machine learning techniques, highlighting their strengths and weaknesses.   This work will help students understand the use of machine learning techniques for practical applications.	40%
Qualifying Condition(s)    A weighted aggregate mark of 40% is required to pass the module.

Machine learning techniques offer a way in which a computational model can be applied to solve complex problems through the suitable training on examples.  These techniques have been used for a wide range of applications, from the simulation of psychological abilities to predicting stock market prices.  The module will examine the biological background of neural computing techniques, provide a foundation on the common approaches to modelling neurons, investigate advanced machine learning techniques, and explore example algorithms for supervised and unsupervised machine learning.

None.

The aim of this module is to equip students with a good working knowledge of the most popular machine learning techniques so that they can apply such techniques with confidence to appropriate problems and describe their benefits and limitations.  Students should be prepared to use basic mathematics (including geometry and calculus) to understand how machine learning  techniques are derived and applied.

By the end of the module, students should be able to:

·         Explain the origin of neural computing techniques in relation to formal models of biological systems.

·         Recognise the type of problems that benefit from the application of machine learning techniques.

·         Distinguish between systems that learn by using different types of learning algorithm.

·         Describe in detail supervised learning techniques and example algorithms.

·         Describe in detail unsupervised learning techniques and example algorithms.

·         Apply machine learning techniques to solve complex problems.

Evaluate the use of machine learning techniques, highlighting their strengths and weaknesses.

The module is divided into the following areas:

·         Natural & Artificial neurons: terminology, principles and early techniques.

·         Input and output data: data selection and preparation

·         Performance measurement techniques

·         Supervised learning systems:

o        Perceptrons,

o        Backpropagation,

o        Bayesian neural networks,

o        Radial Basis Function networks

o        Support Vector Machines

·         Unsupervised learning systems,

o        Hebbian learning,

o        Kohonen’s Self-organising Feature Maps.

Enhancing performance using ensembles of machine learning systems

30 contact hours in weeks 1-10, consisting of a mixture of lecture and example classes, and lab sessions.

Required Reading :

Picton, P. (2000).  Neural Networks, 2nd Edition.  Basingstoke , UK.: Palgrave.

Recommended Reading :

Haykin, S. (2008).  Neural Networks and Learning Machines, 3^rd Edition.  Upper Saddle River , NJ.: Prentice-Hall Inc.

Cristianini, N and Shawe-Taylor (2000), J. An Introduction to Support Vector Machines and other kernel-based learning methods. Cambridge University Press, .

Callan, R. (1999).  The Essence of Neural Networks.  Hemel Hempstead, UK.: Prentice-Hall Europe .

Kohonen, T. (1997).  Self-Organizing Maps, 2nd Edition.  Berlin, Heidelberg, New York : Springer-Verlag.

Other Relevant Sources:

Bishop, C.M. (1995).  Neural Networks for Pattern Recognition.  Oxford, : Clarendon Press.

Arbib, M.A. (Ed) (2003).  The Handbook of Brain Theory and Neural Networks, 2^nd Edition.  Cambridge , MA.: MIT Press.

Journals:

IEEE Transactions on Neural Networks: http://ieeexplore.ieee.org/xpl/RecentIssue.jsp?puNumber=72.

Neural Networks: http://www.sciencedirect.com/science/journal/08936080.

Neurocomputing: http://www.sciencedirect.com/science/journal/09252312.

20 August 2008