Back to list of Invited Talks

Statistical Learning in Multimedia

Sankar Basu,
IBM T. J. Watson Research Center, US

This talk will be in two parts. In the first part we will review some standard tools from statistical learning theory that has found applications in recent research in multimedia search and retrieval. These include the support vector machine formalism, graphical models such as the hidden Markov models, bayesian networks and their generalizations. Use of other learning tools such as the decision trees and neural networks will also be touched upon. In the second part of the talk we will discuss how the above tools have been used in conjunction with low level image and video features to successfully build a generic system for video retrieval. The system makes use of multi-modal data, namely audio and video data for this purpose. Different fusion strategies that can be deployed in this context of research will also be pointed out.

Sankar Basu was born in Calcutta, India, and received Ph.d degree from the University of Pittsburgh, USA. He has been an Alexander von Humboldt fellow on extended visits to Ruhr University, Bochum, Germany over many years, was a visiting scientist at the laboratory for Informations and Decision Systems (LIDS) at MIT, and has been a faculty member of electrical engineering at the Stevens Institute of Technology, where he taught and conducted funded research (US Air-Force, NSF) in signals processing, and was the recipient of 1991 Jess Davis memorial award for excellence in faculty research. For the last several years he has been with the IBM T. J. Watson Research Center, New York.

Dr. Basu's main research interests had been in the mathematics of networks and systems theory with particular emphasis on multidimensional systems. More recently, he has been involved with statistical learning and multimedia signal processing. He has also published in digital filter synthesis, image processing, nonlinear modeling techniques and wavelets and filter banks.

Back to list of Invited Talks

Modeling and Manipulating Large Scientific Datasets

R.Daniel Bergeron,
University of New Hampshire, US

As the size and complexity of modern scientific data sets continues to grow dramatically, it is clear that we need to develop new models and innovative tools for manipulating such data, especially in the context of an interactive data exploration environment. These datasets provide a variety of challenges that simply cannot be met by traditional database technology. The principal problems include:

  1. the implicit relationships that exist among the individual data points in a scientific dataset cannot be effectively or efficiently represented using today's data models;
  2. interactive exploration of these huge datasets requires multiresolution representations that are also not well described with existing data models;
  3. the complex nature of scientific data and the variety of ways in which diverse sources of data can be combined into a single logical dataset cannot be effectively represented in traditional data models.

We need a formal framework for modeling the complex scientific data at multiple levels of resolution. This model must be as much general as possible in order to address a wide range of application domains in a single comprehensive fashion.

This talk will review recent developments in the areas of multiresolution data representation, the representation and modeling of large complex scientific datasets, and scientific database systems. In addition it will describe our own efforts to define a formal data model for multiresolution scientific data and a prototype system to support that model.

R. Daniel Bergeron (Professor, Computer Science) has been a faculty member at the University of New Hampshire since 1974 and is currently chair of the department.

He was founding Editor-in-Chief of ACM Transactions on Graphics and served in that role for six years. He is on the Editorial Board for "Computers and Graphics". He was Program Committee Chair for SIGGRAPH '82, Panels Co-Chair for Visualization '90 and Visualization '91, Program Co-Chair for Visualization '92, Papers Co-Chair for Visualization '93 and Visualization '94 and has served on numerous other program committees for SIGGRAPH, Visualization conferences, and Eurographics. He served on the Executive Board for Eurographics from 1991-2000.

Professor Bergeron has been active in computer graphics since 1967, but his Ph.D. dissertation (Brown University, 1973) was in the area of software engineering. He has research interests and publications in the areas of computer graphics, scientific visualization, scientific database systems, parallel and distributed computing, user interface design, intelligent tutoring systems, software engineering, and expert systems. Professor Bergeron's current research is focused on the integration of multi-dimensional scientific data visualization techniques into a comprehensive scientific database environment.

Back to list of Invited Talks

Techniques for Conformance Testing of Timed Systems

Hacene Fouchal,
Universite de Reims Champagne-Ardenne, F

In the protocol engineering area, the software validation may take two research directions: the verification aspect (proves the correctness of the system) or the testing aspect (finds incorrectness on the implementation). The time factor plays an important role in some new system developments (multimedia protocols, Real-time systems, ...).

In the last five years, some research teams have began to investigate testing systems having strong timing constraints. Many aspects have been tackled: Timed system modelling, Exhaustive testing, Test purpose based techniques.

Up to now, testing timed systems is still a new field. It does not exist any technique better than others. All of them have some restrictions on systems which makes them unrealistic. In any case, many issues have to be investigated: Efficient models, Data integration, Real case studies, Interoperability testing, Distributed system testing.

The aim of the talk is to give some solutions for some of these open problems especially about modelling and test case generation methods.

Back to list of Invited Talks

The evolution of types and functions in logic and computer science

Fairouz Kamareddine,
Heriot-Watt University, Edinburgh, UK

Functions and types play a central role in logic and computation. In this talk, we outline the evolution and impact of types and functions throughout the 20th century.

We describe how the notions of Functionalisation (the way in which functions can be constructed) and Instantiation (the process of applying a function to an argument) were implemented in Frege's Begriffschrift, Russell's Ramified Type Theory and the lambda-calculus showing that the lambda-calculus misses a crucial aspect of functionalisation. We then pay attention to some special forms of function abstraction that do not exist in the lambda-calculus and we show that various logical constructs used in computational systems (e.g., let expressions, contexts and parameters) can be seen as forms of the missing part of functionalisation. Our study leads to a nice refinement of well known type systems which enable a natural and more accurate representation of computational systems like de Bruijn's famous Automath, the Logical Framework LF and the functional language ML.

This is joint work with Twan Laan and Rob Nederpelt.

Back to list of Invited Talks

Equivalence-Checking and Model-Checking with Infinite State Systems

Tony Kucera,
Masaryk University Brno

A joint work with

Petr Jancar,
Technical University Ostrava

Two dominating approaches to formal verification of concurrent systems are equivalence-checking and model-checking. In the equivalence-checking approach, one describes the specification and the actual implementation of a system as states in a certain transition system, and then it is shown that the two states are equivalent. In the model-checking approach, desired properties of the implementation are expressed as a formula of a suitable modal logic, and then it is shown that the implementation satisfies the formula.

In the case of finite-state systems, both techniques can be fully automated. Real-world concurrent systems (such as parallel programs) often have infinitely many states. The reason is that they, e.g., operate on infinite data structures, can initiate an unbounded number of new processes during the computation, etc. In the last decade, questions about principal frontiers of automated verification with infinite-state systems attracted a lot of attention. The paper presents the state-of-the-art overview of existing results with emphasis on underlying principals and proof techniques.

Antonin Kucera is an associate professor of computer science at the Faculty of Informatics, Masaryk University in Brno. He obtained his PhD in 1997, and then he spent almost two years at the Technical University Munich as an Alexander von Humboldt Fellow. His research interest are mainly in the area of decidability and complexity issues for infinite-state systems. In the year 1997 he was awarded the Talent'97 prize by the Czech ministry of education. He has (co)authored about thirty papers published in proceedings of recognized conferences and journals.

Back to list of Invited Talks

Is there a Future for University Education (in Computer Science as in Other Disciplines) ?

Elie Milgrom,
Universite catholique de Louvain, BE

This position paper will argue that it is urgent that universities, professional societies, and politicians concerned with higher education devote less attention to curriculum content, European harmonization and quality control, and more to improving the quality of education.

Rather than merely providing opportunities for lecturers to display their mastery of various subjects in front of captive audiences, shouldn't higher education make certain that students do learn what they are expected to learn ?

A rational approach should thus look at the learning process and its outcomes, and implement strategies and mechanisms for achieving desired outcomes as well as for preventing undesired ones. Surprisingly enough, very little from what is known about effective teaching methods seems currently to be put into daily practice in many (most ?) university departments.

The talk will present a number of principles and approaches which effectively contribute to improving the quality and effectiveness of higher education, as well as some practical experiments based on these ideas.

Presentation in PDF

Elie Milgrom graduated in 1967 with a degree of electrical engineering at the University of Ghent (Belgium). He obtained a PhD in Electrical Engineering in 1971 from the Technion - Israel Institute of Technology. He started his academic career as assistant professor at the Courant Institute of Mathematical Science, New York University. In 1972, he was appointed at the School of Engineering of the Université catholique de Louvain with the specific mission to develop research and teaching activities in computing science and engineering. His research interests have ranged from software engineering to distributed systems. Since 1998, he has devoted a significant amount of time and energy to the improvement of the quality of university education. As such, he is one of the originators of a new engineering curriculum at UCL, based on active learning techniques.

Back to list of Invited Talks

Video Indexing and Retrieval by Content

Alberto del Bimbo,
Universita di Firenzi, IT

Development of systems supporting effective indexing and retrieval by content of videos requires to perform a wide spectrum of operations on video streams, including temporal segmentation, analysis of the audio and video tracks, identification and recognition of text. Low level features are then processed to provide some higher level description of video content, as most of the user queries are typically related to higher level syntax and semantics, rather on the lower level lexical level. Moreover, the specificity of different application domains entails that different solutions be adopted in different contexts. This may affect both the choice of low level features to be extracted, as well as the modeling of specific domain knowledge required address the issue of higher level of semantics. In this paper we will report on our experience in the application con texts of news and sports videos. We will show solutions adopted to cope with specifc requirements of different application domains.

Alberto Del Bimbo is Full Professor of Computer Engineering at the Univer sity of Florence, Italy. Since 1998 he is the Director of the Master in Multimedia of the University of Florence. At the p resent time, he is also Deputy Rector of the University of Florence, in charge of Research and Innovation Transfer. His scientific interests are Pattern Recognition, Image Databases, Multimedia an d Human Computer Interaction. He has delved into object recognition and image sequence analysis, content based re trieval, natural human computer interaction. Prof. Del Bimbo is the author of over 170 publications in internat ional journals and conference proceedings. He is the author of the "Visual Information Retrieval" monography, Morgan Khaufman Ed, 1999. Prof. Del Bimbo was the President of the IAPR Italian Chapter, from 1996 to 2000 and M ember at large of the IEEE Publication Board from 1998 to 2000. He is Associate Editor of Pattern Recognit ion, Journal of Visual Languages and Computing, Multimedia Tools and Applications Journal,Pattern Analysis and Applications, IEEE Transactions on Multimedia, and IEEE Transactions on Pattern Analysis and Machine Intelligenc e. He was the Guest Editor of many special issues in highly respected jpurnals. He was the General Chairman o f the 9th IAPR International Conference on Image Analysis and Processing, ICIAP '97 - Florence (1997) and the General Chairman of the 6th IEEE International Conference on Multimedia Computing and Systems, ICMCS'99 - Fl orence (1999).

Back to list of Invited Talks

Interactive Indexing and Retrieval of Multimedia Content

Marcel Worring,
University of Amsterdam, NL

Retrieving information from a multimedia database is difficult due to the semantic gap between the user's perception of the data and the descriptions we can derive automatically from the data using computer vision, speech recognition, and natural language processing. In this presentation we consider two classes of methods which help in narrowing or briding the gap: The talk will consist of a review of the state-of-the-art as well as examples from our own research.

Marcel Worring received his MSc. degree (honors) in computer science from the Free University Amsterdam ('88) and his PhD in '93 from the University of Amsterdam (UvA). Currently he is an assistant professor at the UvA focussing at multimedia information analysis. In this field he leads a large multidisciplinary project. He authored and co-authored over 70 publications, many of them in leading journals or conferences. Next to this he is co-founder of MediaMill, an application centre for multimedia solutions.

Back to list of Invited Talks

A Computer Scientist's View of Wireless Communications

Stephan Olariu,
Old Dominion University, US

The main goal of this address is to present, from a computer scientist's perspective, the key issues in the emerging area of mobile computing enabled by unprecedented technical advances in wireless communications.

The talk will provide a survey of a number of major accomplishments in the area of mobile computing and wireless network. In addition, important new trends will in both terrestrial and satellite networks will be discussed and analyzed.

Back to list of Invited Talks

Parameterised Contracts and Rich Interface Descriptions for Software Components

Ralf Reussner,
Monash University, AU

This talk paper presents an approach for predicting functional and extra-functional properties of layered software component architectures. Our approach is based on parameterised contracts, a generalisation of design-by-contract. Firstly, the talks adresses the meaning of "contractual use of components" a term sometimes used loosely -- or even inconsistently -- in current literature. Secondly, we introduce the abstract concept of parameterised contracts. It is shown how parameterised contracts are used to compute the properties of a component in dependence from its environment.

In particular, this talk concentrates on describing component protocols and their dependence from the components environment. Component interface models, specifying in addition to signature-lists also information about components protocols, etc. are often called rich interfaces. We describe an interface model containing an extension of finite state machines is described which

This enables us to perform interoperability and substitutability checks statically. Parameterised contracts add the possibility to perform a certain class of protocol adaptations automatically.

Ralf H. Reussner received his diploma in computer science in 1997 from the Universtaet Karlsruhe (T.H.), Germany. He was member of the software engineering group at the University of Kaiserslautern 1997--1998. From 1998 to 2001 he received a research fellowship from the National German Research Council (DFG) and received his PhD in computer science in 2001 from the Universtaet Karlsruhe (T.H.). In his Phd-thesis he introduced the abstract concept of parameterised contracts and worked on extensions of finite state machines having a decidable inclusion problem. Since then he has been a Senior Research Scientist at the Distributed Systems Technology Centre Pty Ltd. (DSTC), Melbourne, Australia, where he works on mechanisms for predicting properties of component-based software architectures. He published a book and about thirty articles on software engineering for distributed and parallel applications in journals and on conferences, in particular about benchmarking, component interface models and automatic component adaption. Related to this research interests, he set up and maintains the opensource project SKaMPI. He consults in the area of distributed applications. He held several invited talks about his research in academia and industry. Besides serving on several program committees, he organised workshops on on ECOOP and TOOLS conferences. He is member of the ACM, IEEE, and the GI e.V.

Back to list of Invited Talks