Professor Jose Aguilar
CEMISID. Dpto. de Computacion
Facultad de Ingenieria
Universidad de los Andes
Av. Tulio Febres. Merida
Venezuela
E-mail: aguilar@ula.ve

Abstract: In this plenary we present recent extensions to the Fuzzy Cognitive Maps (FCM) to improve their performances (learning procedures, FCM hierarchical model, dynamical FCM, etc.). Additionally, we present several applications in different domains (social systems, control systems, multiagent systems, etc.). This technique is the fusion of the advances of the Fuzzy Logic and Cognitive Maps theories. FCMs were proposed by Kosko to represent the causal relationship between concepts and analyze inference patterns [10, 11, 12, 13]. FCMs represent knowledge in a symbolic manner and relate states, processes, policies, events, values and inputs in an analogous manner. Once constructed, an FCM allows performing a qualitative simulation of the system and experiment with the model. FCMs have gradually emerged as a powerful modelling and simulation technique applicable to numerous research and application fields: administrative sciences, game theory, information analysis, popular political developments, electrical circuits analysis, cooperative man–machines, distributed group-decision support, etc.
Some application examples are presented at the following. [6, 19, 21, 22] investigate the implementation of the FCM in control problems. Particularly, FCMs have been used to model and support a plant control system, to construct a system for failure modes and effect analysis, to model the supervisor of a control system or of manufacturing systems. FCMs have been used in multiagents system to represent different types of knowledge in a group of agents, to support the building of group consensus, like ontological framework to share knowledge [9, 15, 20]. FCM has been used to structure virtual worlds that change with time [7, 11]. FCM links causal events, actors, goals, and trends in a fuzzy feedback dynamical system. It can guide actors in a virtual world as the actors move through a web of cause and effect and react to events and to other actors.
To improve the performance of FCM, several works have been proposed. For example, three core issues are discussed and respective solutions are proposed in [14]: the first one concerns the case of multi-stimulus situations (parallel stimulation of many FCM concepts), the second one focuses on the design of a learning algorithm (using evolution strategies), and finally the generic real-world phenomena of conditional effects and synergies are properly modelled to support the inference mechanism of FCMs. Carvalho et al. propose a Rule Based Fuzzy Cognitive Map (RBFCM), as an evolution of Fuzzy Causal Maps (FCM) that allow a more complete representation of cognition, since relations other than monotonic causality are made possible [3, 4].
The purpose [2] is to describe a FCM based on the random neural network model called the Random Fuzzy Cognitive Map (RFCM). This model is based on the probability of activation of the neurons/concepts in the network. The model carries out inferences via numerical calculation instead of symbolic deduction. In [1] is described an Adaptive Random Fuzzy Cognitive Map (ARFCM). The ARFCM changes its fuzzy causal web as causal patterns change and as experts update their causal knowledge. They show how the ARFCM can reveal implications of models composed of dynamic processes. [5, 6] have proposed a dynamical FCM (DFCM) to implement Causal Relations like adjustment functions. The adjustment functions can be based on fuzzy rules or mathematical equations of the modelled system.
In general, the task of creating FCMs is made by experts in a certain domain, but is very promising the automatic creation of FCMs form raw data. In [23] Vazquez presents a new algorithm (the Balanced Differential Algorithm) to learn FCMs from data. To enable the gradual learning of symbolic representations, a backpropagation learning procedure has been developed for FCM [16]. In [17, 18] have been proposed FCM hierarchical models and unsupervised learning techniques for tuning FCMs.
Several tools based on FCMs have been developed for different problems. The FCModeler tool displays the known and uncertain biological information in a metabolic network using interactive graph visualization [8]. The system also models pathway interactions and the effects of assumptions using a FCM-based modelling tool. We have proposed a tool to implement DFCM in [5].

Brief Biography of the Speaker:
Professor Jose Aguilar received the B. S. degree in System Engineering in 1987 from the Universidad de los Andes-Merida-Venezuela, the M. Sc. degree in Computer Sciences in 1991 from the Universite Paul Sabatier-Toulouse-France, and the Ph. D degree in Computer Sciences in 1995 from the Universite Rene Descartes-Paris-France. He was a Postdoctoral Research Fellow in the Department of Computer Sciences at the University of Houston (1999-2000). He is a Titular Professor in the Department of Computer Science at the Universidad de los Andes (ULA), researcher of the Center of Studies in Microelectronics and Distributed Systems (CEMISID). Currently, he is head of the Free Technology Research Center (CENDITEL). He was head of the Science and Technology Bureau of the Merida State, Venezuela, during 6 years (2001-2007), coordinator of CEMISID from 2001 to 2007, and belonged to the committee that created the High Performance Computing Center of the ULA (CeCalCULA) in 1995.
He has published more than 200 papers in journals, books and proceedings of international conferences in the field of parallel and distributed systems (performance evaluation, task/data/transaction assignment and scheduling, fault tolerance, middleware design, etc.), computational intelligence (artificial neural networks, evolutionary computation, fuzzy logic, swarm intelligence, multiagente systems, etc.) applied to combinatorial optimization, pattern recognition, control systems (identification and supervision systems, distributed and intelligent control, industrial automation, etc.), among others. He has published 5 books in the domain of computational sciences, and science and technology management. He has been Chairman of Symposia, Workshops, etc.; editor of proceedings and books, and member of more than 30 Program Committees for different International Conference and scientific juries. He has more than 50 conferences in different international or national congress. In addition, he has participated in training courses both nationally and internationally. He has received several awards and some of his papers have received special awards. The last 6 years has been one of the two best researchers of his university (The University has more than 1000 researchers). Dr. Aguilar has been a visiting research/professor in different universities and laboratories (Universite Pierre et Marie Curie-Paris-France, Universite de Versailles Paris-France, Universite Rene Descarte-Paris-France, Laboratorie d’Automatique et Analyses de Systemes-Toulouse-France, University of Houston-USA, Universidad de la Coruna-Spain, Universidad Complutense de Madrid-Spain, Institute National de Recherche en Informatique Niza-France). He has been the coordinator or inviting research in more than 20 research or industrial projects supported by the Venezuelan Scientific Office, the French Scientific Office, the Scientific Office of the Universidad de los Andes, INTEVEP (Venezuelan Institute of research in oil), the European Economic Community, among others. In these projects, he has written more than 40 technical reports. Aguilar has been a consultant for PDVSA (the Venezuelan Oil Company), SIDOR (the Venezuelan Iron and Steel Industry), Venezuelan government departments, etc. Aguilar has supervised more than 25 M.S. and Doctoral students in their thesis and dissertation work. He is currently supervising 5 Ph.D. dissertations and 2 M.S. thesis.