T4: Rule Base Compression in Fuzzy Systems

Description

The tutorial presents a systematic study on the inherent complexity in fuzzy systems, resulting from the large number and the poor transparency of the fuzzy rules. The study uses a novel approach for complexity management that compresses the fuzzy rule base by removing the redundancy and preserving its behaviour. The compression is based on methods for presentation, manipulation, transformation and simplification of fuzzy rule bases, which are illustrated by algorithms as well as results from numerous examples and two case studies. The results are directly applicable or easily extendable to a wide class of fuzzy systems and detailed benchmarks for expanding these systems to new areas such as fuzzy networks and fuzzy multi-agent systems are introduced. The intended participants are people from both academia and industry, who would be interested in building and implementing advanced fuzzy systems.

The tutorial is organised in 10 sections. Section 1 is an introduction to the topic of complexity in general and in the context of fuzzy logic. Section 2 discusses the most common types of fuzzy rule based systems and analyses their impact on complexity. Section 3 reviews most of the existing rule base reduction methods for fuzzy systems and summarises their attributes. Section 4 introduces advanced techniques for presentation of fuzzy systems based on boolean matrices and binary relations, which facilitate the overall management of complexity. Sections 5 and 6 present techniques for manipulation of single rule base fuzzy systems with general and special rule bases, which reduce the qualitative complexity. Sections 7 and 8 describe techniques for transformation of multiple rule base fuzzy systems with feedforward and feedback interconnections, which also reduce the qualitative complexity. Section 9 proposes techniques for simplification of fuzzy rule based systems, which reduce the quantitative complexity by aggregation of inconsistent rules and filtration of non-monotonic rules. The last Section 10 is a conclusion highlighting the theoretical significance of the methodology for complexity management in fuzzy systems, an application framework for this methodology and some future research directions.

About the presenter

Dr Alexander Gegov is Senior Lecturer in the School of Computing at the University of Portsmouth. He holds a PhD in Control Systems and a DSc in Intelligent Systems. His main research interests are in the theory of computational intelligence (fuzzy logic, neural networks and genetic algorithms) and complex systems (hierarchical, decentralized and multilayer systems) as well as their application for modelling, simulation and control in areas such as transport networks and the environment. He has published his main research results in complex systems in a number of international journals such as the International Journal of Control and Systems & Control Letters. He is also the sole author of two books – the first one published in the Kluwer Series in Intelligent Technologies in 1996 and the second one published in the Springer Series in Fuzziness and Soft Computing in 2007. In recent years he has regularly reviewed papers submitted to a number of journals in computational intelligence, such as IEEE Transactions on Fuzzy Sets and Systems and the International Journal of Fuzzy Sets and Systems as well as research proposals submitted to the Australian Research Council. He was a first prize winner for young researchers of the Bulgarian Union of Scientists in 1996, an invited lecturer to the NATO Advanced Study Institute on Soft Computing in 1997 and a presenter at the House of Commons Conference on Promoting Britain’s Young Researchers in 2000. He is an Affiliate of the International Federation of Automatic Control (IFAC) and a Member of the European Society for Fuzzy Logic and Technology (EUSFLAT).