Bill Wilson's Neural Networks Projects
Artificial Intelligence Research Group
School of Computer Science and Engineering
University of New South Wales
My current research interests are here.
Recurrent Network Architectures (1991-1995)
This project began as an attempt to teach to neural networks the graphotactic
patterns of English words: e.g. that a word can end but not begin with "nd".
It developed into a comparison of different recurrent network archictectures
for studying problems of this type, where there the possible value of the next
item (to be predicted by the network) is constrained by the values of the
last few inputs (letters) but not by inputs a long time in the past. For
recurrent network publications.
Tensor Product Networks (1991-)
Human Analogical Problem Solving
Applications of formal systems and neural networks in cognitive modelling.
This project, in collaboration with Graeme Halford, Department of Psychology,
University of Queensland, concerns analogical reasoning and memory processes.
A current direction in this work is to use tensor product networks to model
capacity and analogical reasoning, and relational processing in humans. For
details, see tensor
product network publications.
For simple code to solve a proportional analogy problem, see
C code for
original STAR program. STAR stood for Structured Tensor Analogical
Reasoning, as I recall.
Other/Past Projects on Neural Networks (1987-1991)
These included work with Luyuan Fang and Tao Li on optimization neural networks,
with Agus Rahardja and Arcot Sowmya on face recognition using feedforward
networks, and on handwritten character recognition using a conventional
front-end to preprocess the image, then a NN recognizer. For details, see
optimization / face recognition / character recognition publications.
Neural Networks Research in the UNSW AI and IE Department
Neural networks group home page
Maintained by Bill Wilson,
Bill Wilson's contact info
UNSW's CRICOS Provider No. is 00098G