Plenary Sessions

Time: 8:00 - 9:00
Venue: Convention Hall

June 2 (Monday)
Jim Bezdek, University of West Florida, Pensacola, Florida, USA

June 3 (Tuesday)
David B. Fogel, Natural Selection, Inc., San Diego, California, USA

June 4 (Wednesday)
Teuvo Kohonen, Helsinki University of Technology, Helsinki, Finland

June 5 (Thursday)
Takeshi Yamakawa, Kyushu Institute of Technology, Kitakyushu, Kyushu, Japan

June 6 (Friday)
Christopher M. Bishop, Microsoft Research Cambridge, Cambridge, UK
 

[ Return to top ]

Jim Bezdek
University of West Florida
Pensacola, FL, USA

Visual Cluster Validity

Abstract

Visual cluster validity began in 1939 with the method of correlation profiles given by R. C. Tryon. Since then, there has been a more or less continuous evolution of different visual methods for the validation of clusters. This talk begins with some preliminary information about cluster validity, and then I will give a very short history of visual cluster validity. Emphasis is placed on methods that reorder and image a relational version of the data. This form of validation began with Burt in 1940, and continues to see improvements and applications to the present. Then I will review VAT, sVAT and coVAT, a family of algorithms that build visual representions of square and rectangular relation data. Finally, I will present a new method of visual cluster validity (for the square case) based on (human) comparison of a pair of reordered relation matrices, D* and D (U*). D* is built by the VAT algorithm (Hathaway and Bezdek, 2002); and D(U*) is built by applying a transformation to any crisp or fuzzy partition of the data. (Huband and Bezdek, 2008). I will give some examples of the technique, after showing that this method, like all validation methods, is certain to fail at some point. 

Biography

Jim received the PhD in Applied Mathematics from Cornell University in 1973. Jim is past president of three professional societies: NAFIPS (North American Fuzzy Information Processing Society); IFSA (International Fuzzy Systems Association); and the IEEE Computational Intelligence Society. Jim is the founding editor of two journals: the Int'l. Jo. Approximate Reasoning; and the IEEE Transactions on Fuzzy Systems. Jim is a fellow of the IEEE and IFSA, and recipient of the IEEE 3rd Millennium, IEEE CIS Fuzzy Systems Pioneer, and IEEE CIS Rosenblatt medals. Jim's interests include woodworking, optimization, motorcycles, pattern recognition, cigars, fishing, image processing, blues music, and cluster analysis.

 

[ Return to top ]

David B. Fogel
Natural Selection, Inc.
San Diego, California, USA

The Burden of Proof - Part II

Abstract

Standards of evidence in scientific work, by the very term "standards," should be consistent, but they are not. Often, well-known "facts" or claims turn out to be wrong, disagreements over the interpretation of data and methods yield to political motivations. Even people who would have us strive for the highest aspirations of scientific quality defend arguments from vox populi, or at least majority rule.  This lecture will discuss the standards of evidence in scientific work, with particular emphasis on evolutionary computation and modeling complex adaptive systems. Evidence shows that some seemingly simple systems are really quite complicated. In other cases, adjusting assumptions about a model leads to results that are at significant variance from what is commonly accepted. The implications of accepting well-known models of these systems are explored. Two common concepts are identified as being associated with potential problematic models: expectation and equilibrium.

Biography

Dr. David Fogel is president and CEO of Natural Selection, Inc. in San Diego, California, USA. He received the Ph.D. in engineering sciences (systems science) from UCSD in 1992. Dr. Fogel is president of the IEEE Computational Intelligence Society (2008-2009). He is the author of over 200 publications in scientific literature, including 6 books, and he holds 4 U.S. patents. Dr. Fogel is a Fellow of the IEEE, and the recipient of the 2004 IEEE Kiyo Tomiyasu Technical Field Award and the 2008 IEEE Computational Intelligence Society Evolutionary Computation Pioneer Award. He was the founding editor-in-chief of the IEEE Transactions on Evolutionary Computation.

[ Return to top ]

Teuvo Kohonen
Helsinki University of Technology
Helsinki, Finland

Data Management by Self-Organizing Maps

Abstract

The self-organizing map (SOM) is an automatic data-analysis method. It is widely applied to clustering problems and data exploration in industry, finance, natural sciences, and linguistics. The most extensive applications, exemplified in this paper, can be found in the management of massive textual data bases. The SOM is related to the classical vector quantization (VQ), which is used extensively in digital signal processing and transmission. Like in VQ, the SOM represents a distribution of input data items using a finite set of models. In the SOM, however, these models are automatically associated with the nodes of a regular (usually two-dimensional) grid in an ordered fashion such that more similar models become automatically associated with nodes that are adjacent in the grid, whereas the less similar models
are situated farther away from each other in the grid. This organization, a kind of similarity diagram of the models, makes it possible to obtain an insightful view of the global metric relationships of data, especially of high-dimensional data items. If the data items belong to certain predetermined classes, the models (and the nodes) can be calibrated according to these classes.  An unknown input item is then classified according to that node, the model of which is most similar with it in some metric used in the construction of the SOM. A new aspect introduced in this paper is that an input item can even more accurately be represented by a linear mixture of a few best-matching models. This becomes possible by a least-squares fitting procedure where the coefficients in the linear mixture of models are constrained to nonnegative values.

Biography

Teuvo Kohonen was born on 11 July 1934 in Lauritsala, Finland. He obtained a Dipl. Ing. degree in 1957, and a Licentiate of Technology degree in 1960 from Helsinki University of Technology. After receiving  a D. Eng. degree from the same university in 1963 with a thesis on the annihilation of positrons he was appointed as a Professor of Physics at the Helsinki University of Technology. Since 1975 he has also acted as a Research Professor of the Academy of Finland.

Dr. Kohonen pioneered computer science education in Finland, but already in the 1960s he started to pursue research on a field that nowadays might be called "artificial neural networks." He has written five books (in English) and hundreds of scientific papers on digital technology, associative memories, pattern recognition, and self organization. Most well known is his data-analysis method called the Self-Organizing Map (SOM) that produces low-dimensional graphic representations of the similarity relationships in high-dimensional data sets.
This method has been adopted widely for the visualization and adjustment of industrial processes, and  at large, for data exploration in finance, trade, natural sciences, and linguistics. Some 8000 scientific papers and a dozen of books have been written on Kohonen's SOM method, and six international workshops have been organized on the SOM.

Dr. Kohonen is a Fellow of the IEEE. He has received a number of awards, including the IEEE Neural Networks Pioneer Award, the Technical Achievement Award of the IEEE Signal Processing Society, the INNS Lifetime Achievement Award, the King-Sun Fu Prize of the International Association for Pattern Pecognition, and many others. He was nominated as an Academician of Finland (a title simultaneously held only by twelve persons), and is a member of four Academies including the Academia Scientiarum et Artium Europaea, and Académie Européenne des Sciences, des Arts et des Lettres. He is a Dr. h.c. of three universities.

 

[ Return to top ]

Takeshi Yamakawa
Kyushu Institute of Technology
Kitakyushu, Kyushu, Japan

Bio-inspired Self-Organizing Relationship Network as Knowledge Acquisition Tool and Fuzzy Inference Engine

Abstract

A human brain facilitates the topological mapping of the external complex information (multiple-dimensional) to the cortex (two-dimensional) by sensing and watching. The topological mapping was modeled as SOM (Self-Organizing Maps) by Teuvo Kohonen (1982), which enables the vector quantization and the reduction of multiple dimension of information to one or two dimensions (visualization of similarities). Since the SOM visualizes, on the competitive layer, the similarity of raw information, it can be utilized in the field of pattern classification, data analysis, and so on. However, it cannot model the input-output characteristics of the system of interest, that is, it cannot represent the knowledge of a human expert.

In order to squeeze out the essence from the data set with evaluation obtained by trial and error, the novel modeling tool was developed by the author (1999), which is the extension of SOM and in which the input-output relationship of the system is mapped onto the competitive layer. The system is named as self-organizing relationship network (SOR network). A set of units on the competitive layer of the SOR network after learning exhibits a set of typical input-output characteristics of the system of interest and thus the network achieves the knowledge acquisition (IF-THEN rules) from the raw data with evaluation.

The evaluation for each data necessary for the learning of the SOR network is possibly intuitive and deterministic. The plenary talk presents the applications of SOR network, the evaluation of which is intuitive or deterministic.

Biography

Prof. Takeshi Yamakawa is a professor of Graduate School of Life Science and Systems Engineering, Kyushu Institute of Technology (KIT), Japan. He established a foundation, Fuzzy Logic Systems Institute (FLSI), in Japan in 1990 in order to promote the international collaboration on fuzzy logic, neural networks and other soft computing, which are the new paradigms to realize the advanced intelligent systems. He is also the chairman of FLSI. His main research interest lies on hardware implementation of fuzzy systems, fuzzy neural networks, chaotic systems, self-organizing systems, design and fabrication of integrated circuits and micro-total-analysis systems.

Prof. Yamakawa plays Karate (Japanese traditional martial arts) and possesses a black belt (5th Dan). And he likes swimming, horse riding and monocycle as well. His interest also lies on Shakuhachi and Shamisen, which are Japanese traditional musical instruments.

[ Return to top ]

Christopher M. Bishop
Microsoft Research Cambridge
Cambridge, UK

A New Framework for Machine Learning

Abstract

The last five years have seen the emergence of a powerful new framework for building sophisticated real-world applications using machine learning. The cornerstones of this approach are (i) the adoption of a Bayesian viewpoint, (ii) the use of graphical models to represent complex probability distributions, and (iii) the development of fast, deterministic inference algorithms, such as variational Bayes and expectation propagation, which provide efficient solutions to inference and learning problems in terms of local message passing algorithms. In this talk I will review the key ideas behind this new framework, and will highlight the benefits compared to traditional methods such as neural networks and support vector machines. The talk will be illustrated with example large-scale applications.

Biography

Chris Bishop is Deputy Director of Microsoft Research Cambridge, and is a Professor of Computer Science at the University of Edinburgh. His first degree was in Physics from Oxford, and he has a PhD from Edinburgh in quantum field theory.

He was the principal organiser of the six month international research programme on "Neural Networks and Machine Learning" at the Isaac Newton Institute for Mathematical Sciences in Cambridge, which attracted several hundred participants. He is the author of the new text book "Pattern Recognition and Machine Learning" which is becoming widely recognised as the leading textbook in the machine learning field. His previous text book "Neural Networks for Pattern Recognition" was also widely adopted.

Chris is a Fellow of Darwin College Cambridge, a Fellow of the Royal Academy of Engineering, and a Fellow of the Royal Society of Edinburgh. His research interests include probabilistic approaches to machine learning, as well as their application to fields such as bioinformatics and medicine. He holds a Commercial Pilot's Licence, and for relaxation he enjoys flying light aircraft, particularly aerobatics.