Alexander Gegov, University of Portsmouth / Technical University of Sofia, alexander.gegov@port.ac.uk, https://www.port.ac.uk/about-us/structure-and-governance/our-people/our-staff/alexander-gegov

Farzad Arabikhan, University of Portsmouth, farzad.arabikhan@port.ac.uk, https://www.port.ac.uk/about-us/structure-and-governance/our-people/our-staff/farzad-arabikhan

Alexandar Ichtev, Technical University of Sofia, ichtev@tu-sofia.bg, https://www.researchgate.net/profile/Alexandar-Ichtev

Tutorial Abstract

The goal of this tutorial is to make participants familiar with fuzzy rule based networks and their inherent interpretability which makes them suitable for building explainable artificial intelligence models. As opposed to current machine learning models which can reflect mainly statistical correlations between input and output variables, fuzzy rule based network models use machine reasoning that allows them to reflect also causal relationships by means of intermediate variables in their network structure. The nodes of fuzzy rule based networks are fuzzy systems represented by rule bases and the connections between these nodes are outputs from and inputs to these rule bases. In this context, apart from being a structural counterpart for a neural network, a fuzzy rule based network is also a conceptual generalisation of a fuzzy system and a bridge between two established types of fuzzy systems- flat and a hierarchical.

Tutorial description

This tutorial has two parts – theory and applications. Each part will finish with a discussion to allow the audience to ask questions and make comments. The material will be presented consistently at a fairly general level whereby theoretical concepts will be introduced mathematically and illustrated visually.

The main goal of the tutorial is to make the conference participants familiar with fuzzy networks and their inherent interpretability which makes them suitable for building explainable artificial intelligence models. These models facilitate the identification of causal relationships between external inputs and outputs by means of intermediate variables.

Fuzzy networks are similar to neural networks in terms of general structure. However, their nodes and connections are different. The nodes of these fuzzy networks are fuzzy systems represented by rule bases and the connections between the nodes are outputs from and inputs to these rule bases. In this context, apart from being a structural counterpart for a neural network, a fuzzy network is also a conceptual generalisation of a fuzzy system as well as bridge between two established types of fuzzy systems – flat and hierarchical.

Fuzzy networks have an underlying two-dimensional grid structure with horizontal levels and vertical layers. The levels represent spatial hierarchy in terms of network breadth and the layers represent temporal hierarchy in terms of network depth.

The nodes of fuzzy networks are modelled by Boolean matrices or binary relations. The connections between the nodes are modelled by block schemes or topological expressions. Each network node is located in a cell within the underlying grid structure.

Nodes in fuzzy networks are manipulated by merging and splitting operations. The merging operations are for network analysis and the splitting operations are for network design. These operations are used for converting a fuzzy network into a fuzzy system and vice versa.

The operations are illustrated on feedforward and feedback fuzzy networks. Feedforward networks include combinations of narrow/wide and shallow/deep network structures. Feedback networks include combinations of single/multiple and local/global feedback loops.

Fuzzy networks are applied to several benchmark examples and validated successfully against flat and hierarchical fuzzy systems. The validation uses performance evaluation indicators for feasibility, accuracy, efficiency, transparency.

In addition to the theoretical concepts above, the tutorial will also present detailed case studies on the application of fuzzy networks for telecommuting modelling and software maintainability prediction. The case studies are benchmarked against flat and hierarchical fuzzy systems.

Tutorial Organisers

Alexander Gegov is Associate Professor in Computational Intelligence in the School of Computing at the University of Portsmouth. He is also Visiting Professor in Control Theory in the English Faculty of Engineering at the Technical University of Sofia. He holds a PhD in Cybernetics and a DSc in Artificial Intelligence – both from the Bulgarian Academy of Sciences. He has been a recipient of a national award for Best Young Researcher from the Bulgarian Union of Scientists. He has been Humboldt Guest Researcher at the Universities of Duisburg and Wuppertal in Germany. He has also been EU Visiting Researcher at the Delft University of Technology in the Netherlands.

Alexander Gegov’s research interests are in the development of artificial intelligence and machine learning methods as well as their application for modelling and simulation of complex systems and networks. He has guest edited several special issues of journals and books with conference proceeding published by IEEE and Springer. He has authored 5 research monographs and more than 20 book chapters published by Springer. He has also authored more than 100 articles and papers in a wide range of peer-reviewed specialised journals and international conferences including IEEE journals and conferences. He has presented more than 20 invited lectures and tutorials at international scientific events including IEEE, EPSRC and NATO Conferences and Summer Schools on Artificial Intelligence, Fuzzy Systems, Neural Networks, Intelligent Systems, Computational Intelligence, Cybernetics and Complexity Science.

Alexander Gegov is Associate Editor for the IEEE Transactions on Artificial Intelligence, IEEE Transactions on Fuzzy Systems, the International Journal of Intelligent Systems, the Journal of Fuzzy Sets and Systems and the International Journal of Computational Intelligence Systems. He is currently Member of the Soft Computing Technical Committee of the IEEE Society of Systems, Man and Cybernetics and the Outstanding Paper Award Committee for IEEE Transactions on Fuzzy Systems of the IEEE Computational Intelligence Society. He is also Member of the IEEE Working Groups on Explainable Artificial Intelligence and Trustworthy Artificial Intelligence as well as the IEEE Task Forces on Explainable Fuzzy Systems and Fuzzy Systems Software.

Farzad Arabikhan joined the University of Portsmouth, UK, as a Lecturer in 2017 after completing his PhD on Modelling Telecommuting using Fuzzy Networks at the same university.  He is now a Senior Lecturer and the course leader of BSc Data Science and Analytics program. He has published his research results in several journal articles and conference papers. He has been involved in number of projects totalling around £22 million mainly from Innovate UK where the focus has been the application of AI in decarbonisation and transport. He has also secured funding from the EU COST Programme for research collaboration with leading academics at the Paris-Sorbonne University, France and the Mediterranean University of Reggio Calabria, Italy. He holds BSc and MSc degrees in Civil Engineering and Transportation Engineering from the Sharif University of Technology, Tehran, Iran.

Alexandar Ichtev is Associate Professor at the Faculty of Automatics, Technical University of Sofia, Bulgaria. He has been staff member of the faculty since 2004. He is currently Head of Systems and Control Group at the same faculty. He was awarded PhD in fault diagnosis and fault tolerant control from Technical University of Sofia in 2004. The focus of the thesis was on neuro-fuzzy models for fault detection and fault tolerant control. Half of the PhD research was conducted at the Delft University of Technology, the Netherlands. He is author of five books and more than 70 articles and papers published in pre-reviewed journals and conference proceedings, including IEEE International Conference on Fuzzy Systems, IEEE International Conference on Intelligent Systems, WSEAS Press, International Journal Informatica and others. He took part in more than 25 scientific research projects both national and international. Also, he was head of two of these projects. He is member of several research networks including CEPUS.