Hybrid neural-fuzzy modeling is an important technology for the development of intelligent systems. Recent developments, especially in control, have shown the benefits of joining these technologies. both technologies provide vehicles for modeling the human thought processes: fuzzy set theory can be seen as being extremely useful at representing the macro-level while the neural technologies provide a unique tool for dealing with the micro-level. Particular benefit will come from a symbiotic relationship in which fuzzy logic structures are used in crafting of neural systems and use the resulting neural networks for the learning and actual implementation. In this spirit the overriding objective of this research is to use the structures provided by fuzzy logic to help guide in the discovery of new neural architectures to better model the sophisticated modes of human reasoning. Motivated by this the basic focus of this project is to investigate some specific fuzzy modeling paradigms and provide for their neural implementation. The first part of the research will look at an extension of the pervasively used fuzzy systems modeling technology, one which allows for a hierarchical representation of the rule base. The goal here is to provide for a neural architecture to implement this modeling and in doing so, among other objectives, get a deeper understanding of the possibilities of representing hierarchical know in a neural framework. The next part will look at the problem of multi-criteria decision making and focus on the neural modeling of complex relationship between criteria. The fist part will investigate a new paradigm for learning called participatory learning which can be seen as a type of critic learning approach.
