Authorised by Academic Registrar, April 1996
Objectives At the completion of this subject students should have in-depth knowledge of selected neural network architectures their theory and applications; be able to develop neural network applications for real-life problems; be able to contrast symbol-based AI with the connectionist approach; and appreciate the current developments in neural networks.
Synopsis The syllabus varies according to current developments in artificial intelligence. For 1996 the theme will be neural networks. The subject covers the connectionist paradigm also known as neural networks. There will be a strong emphasis on the development of neural network applications for business and industry. The following topics will be covered: symbolic AI vs connectionist approach, the history, architecture and biological motivations, learning in neural networks, selected neural network architectures, fuzzy neural networks, extracting rules from neural networks, neural network expert systems, hybrid systems, development of neural network applications, connectionist models of human cognition, current research in neural computing.
Assessment Practical work: 60% + Seminar paper (2500 words): 40%