Plenary Talks
 |
José Luís Balcázar
|
|
In preparation... |
|
AbstractThe Logics of Association Rule Mining Association rules are exact or approximate implications found empirically on relational data. The logic of exact implications has been studied deeply in several disciplines. The study of the logic of partial, or approximate, implications is also advanced, but not to a similar extent at all. Nowadays, the Data Mining field provides ample room for case studies focusing on the goal of extracting implications from data. This contribution belongs to a large research program contemplating the need of further, deeper scientific knowledge about the combinatorics of partial implications. Existing, important progress on various mathematical problems of direct practical relevance for the task of mining associations from data is, currently, somewhat hindered by the lack of deep knowledge about these combinatorics. We propose here an approach based on the most classical, elementary notions of Mathematical Logic: the propositional logic domain, with its traditional semantics in terms of Boolean algebra and entailment, and with its traditional approaches to syntactic calculi for deductive inference. The intuition of entailment can be formulated in a number of ways, but their algorithmic difficulties range from exponential time (co-NP-completeness) up to sheer undecidability. Even purportedly simple cases such as the propositional domains still present formidable difficulties for a number of algorithmic problems. Modeling rationality asks, in fact, for even further expressiveness, but this leads to extremely inefficient algorithmics. There is an intuitively natural choice of a logic of about as much expressivity as possible under the constraint of fast deduction algorithms: Horn logic. Association Rules constitute a very close relative, whose logical, combinatorial, and algorithmic properties are still not fully understood. They can be interpreted as an answer to the following question: how can Propositional Horn Logic, limited in expressiveness but enjoying very efficient entailment tests, help us in the analysis of existing phenomena, given through either interaction with some unknown or only partially known system, or plainly as a static dataset with information gathered about the phenomenon. Many important tasks fall into such a category: scientific discovery from large masses of data gathered by instruments or sensors, decision making in economically crucial environments such as production sectors or forecast in stock market investments and gambling, or processing of information related to human communication like ontologies or social networks. We will describe some recent advances in the topic of inference of implications from data, first, in the active learning setting, where we will explain an important conceptual advance that allows us to simplify the proof of the well-known AFP algorithm and reach a profound understanding of the properties that make it work. Then, we will focus on three problems of inference from static data: the lack of scientific guidance to set to appropriate values the free parameters, such as intensity of implication, inherent to all such data mining algorithms; the problem of constructing a principled approach to process of structured data where the interdependences go beyond the case of plain relational models; and the question of how to react in the case of huge sets of output rules, which is, in practice, almost always. We will describe our Logic-based approach to these problems, and explain in detail how a semantic notion of entailment encompasses the existing notions of redundancy in Association Rules, how it can be characterized by a syntactic deductive calculus, and how these progresses allow us to construct axiomatizations of partial association rules that can be proved to be of absolutely minimum size with respect to the corresponding notion of redundancy. |
|
 |
Stefano Cagnoni
|
|
Stefano Cagnoni graduated in Electronic Engineering at the University ofFlorence in 1988 where he has been a PhD student and a post-doc until1997, working in the Bioengineering Lab of the Department of ElectronicEngineering and Telecommunications. He received the PhD degree inBioengineering in 1993. In 1994 he was a visiting scientist at theWhitaker College Biomedical Imaging and Computation Laboratory at theMassachusetts Institute of Technology. Since 1997, he has been with theDepartment of Computer Engineering of the University of Parma, where he iscurrently Associate Professor. Stefano Cagnoni's main research interests are in the fields of Computervision, Robotics, Evolutionary Computation and Neural Networks. He iseditor-in-chief of the Journal of Artificial Evolution and Applications,chair of EvoIASP, the European Workshop on applications of EvolutionaryComputation to Image Analysis and Signal Processing, and member of theProgramme Committee of several International Conferences and Workshops. Hehas been reviewer for over 15 journals. |
|
AbstractEvolutionary Computation and Computer Vision: algorithm hybridization and new approaches
|
|
 |
Andries Engelbrecht
|
|
In preparation... |
|
AbstractCIlib: A Component-based Framework for Plug-and-Simulate Hybrid Computational Intelligence Systems
|
|
 |
Imre Rudas
|
|
Imre J. Rudas graduated from Bánki Donát Polytechnic, Budapest in 1971, received the Master Degree in Mathematics from the Eötvös Loránd University, Budapest, the Ph.D. in Robotics from the Hungarian Academy of Sciences in 1987, while the Doctor of Science degree from the Hungarian Academy of Sciences. He received his first Doctor Honoris Causa degree from the Technical University of Kosice, Slovakia and his second Honorary Doctorate from University Polytechnica Timisoara, Romania. He is active as a full university professor and Head of Department of Intelligent Engineering Systems. He serves as the Rector of Budapest Tech from August 1, 2003 for a period of four years. He is a Fellow of IEEE, Senior Administrative Committee member of IEEE Industrial Electronics Society, member of Board of Governors of IEEE SMC Society, Chairman of the Hungarian Chapters of IEEE Computational Intelligence and IEEE Systems, Man and Cybernetics Societies. He is the President of the Hungarian Fuzzy Association and Steering Committee Member of the Hungarian Robotics Association and the John von Neumann Computer Society. He serves as an associate editor of some scientific journals, including IEEE Transactions on Industrial Electronics, member of editorial board of Journal of Advanced Computational Intelligence, member of various national and international scientific committees. He is the founder of the IEEE International Conference Series on Intelligent Engineering Systems and IEEE International Conference on Computational Cybernetics, and some regional symposia. He has served as General Chairman and Program Chairman of numerous scientific international conferences. His present areas of research activity are: Computational Cybernetics, Robotics with special emphasis on Robot Control, Soft Computing, Computed Aided Process Planning, Fuzzy Control and Fuzzy Sets. He has published one book, more than 400 papers in books, various scientific journals and international conference proceedings. |
|
AbstractWhen one considers fuzzy subsets of a universe, in order to generalize the Boolean set-theoretical operations like intersection, union and complement, it is quite natural to use interpretations of logic connectives 'and', 'or' and 'not', respectively. It is usually assumed that the conjunction is interpreted by a triangular norm (t-norm for short), the disjunction is interpreted by a triangular conorm (shortly: t-conorm), and the negation by a strong negation. Nowadays it is needless to define t-norms and t-conorms in papers related to theoretical or practical aspects of fuzzy sets and logic: researchers have learned the basics and these notions have become part of their everyday scientific vocabulary. Nevertheless, from time to time it is necessary to summarize recent developments even in such a fundamental subject. This is the main aim of the present paper. Somewhat subjectively, we have selected topics where, on one hand, essential contributions have been made, and on the other hand, both theoreticians and practitioners may find it interesting and useful. First a detailed description of our present knowledge on left-continuous t-norms is presented. Then some new classes of associative operations (uninorms, nullnorms) are reviewed. Finally we demonstrate the role of the evaluation scales (especially the case of totally ordered finite sets) in the choice of operations. |
|
 |
Albert Y. Zomaya
|
|
Albert Y. ZOMAYA currently holds the Chair of High Performance Computing and Networking in the School of Information Technologies at Sydney University. He is the author/co-author of seven books, more than 300 papers, and the editor of eight books and eight conference proceedings. He serves as an associate editor for 16 leading journals. Professor Zomaya is the recipient of the Meritorious Service Award (in 2000) and the Golden Core Recognition (in 2006), both from the IEEE Computer Society. He is a Chartered Engineer (CEng), a Fellow of the American Association for the Advancement of Science, the IEEE, the Institution of Engineering and Technology (U.K.), and a Distinguished Engineer of the ACM. |
|
AbstractHybrid Intelligent Methods for Solving Mobility Management Problems The talk will address some of the key algorithmic and computational challenges associated with the mobility management problem. The talk will present several scenarios for static and dynamic mobility management instances incorporating a combination of metaheuristics. The studies show that hybrid approaches are more capable at producing efficient solutions. From a practical standpoint, these approaches have the potential to lead to massive savings in the number of network signal transactions made to locate users. Several hybrid approaches are used with a number of test networks to show their advantages to the currently implemented GSM standards. The results provide new insights into the mobility management problem. |
|
