Het Wetenschappelijk Interdisciplinair Seminarie
Department of mathematics and computer science, University of Antwerp
Het Wetenschappelijk Interdisciplinair Seminarie
Department of mathematics and computer science, University of Antwerp
11-02-2009
Prof. Dr. Ir. Vincent Rijmen (Katholieke Universiteit Leuven)
10 years of Rijndael
The block cipher Rijndael, which is now the AES, was designed 10 years ago. In this talk, I recall the most important events in the design process and the AES selection process. I also survey the new findings of the post-selection evaluation process, which continues until today. Finally, I compare Rijndael/AES to some recent competitors, namely the stream ciphers that were evaluated in the eSTREAM competition organized by the EU Network of Excellence in Cryptology (ECRYPT).
18-02-2009
Prof. Dr. Geert Molenberghs (Universiteit Hasselt, Katholieke Universiteit Leuven)
Repeated Measures in Medical Statistics
In an example-based fashion, and based on Verbeke and Molenberghs (Springer, 1997, 2000), Molenberghs and Verbeke (2005), and Fitzmaurice, Davidian, Verbeke, and Molenberghs (2009), the concept of longitudinal studies will be introduced. Apart from longitudinal preclinical and clinical trials, other hierarchical, or clustered, data settings will be encountered in other settings, such as the Belgian Health Interview Survey and data from developmental toxicity studies. A brief and informal introduction of the linear mixed model will be given, as a key paradigm for the analysis of (continuous) longitudinal data.
References:
Fitzmaurice, G., Davidian, M., Molenberghs, G., and Verbeke, G. (2009). Longitudinal Data Analysis. Handbooks of Modern Statistical Methods. New York: Chapman & Hall/CRC.
Molenberghs, G. and Verbeke, G. (2005) Models for Discrete Longitudinal Data. New York: Springer.
Verbeke, G. and Molenberghs, G. (1997). Linear Mixed Models in Practice: A SAS Oriented Approach. Lecture Notes in Statistics 126. New York: Springer.
Verbeke, G. and Molenberghs, G. (2000) Linear Mixed Models for Longitudinal Data. New York: Springer.
04-03-2009
Dr. Ir. Filiep Vanpoucke (Universiteit Antwerpen)
Cochlear Implants: biomorphic sound processing and modelling
Cochlear implants represent the most successful brain machine interface. Worldwide more than 150 000 deaf persons have regained their hearing with this auditory prosthesis. The cause of deafness in these persons is sensorineural: the tiny hair cells on the basilar membrane that are - in a healthy ear -responsible for converting mechanical vibrations into neural firing patterns, are destroyed. As a consequence incoming sound is no longer reaching the brain. A cochlear implant bridges this gap by surgical placement of an electrode carrier in the inner ear that can connect to the auditory nerve by means of externally electrical pulses, effectively bridging the gap.
In its short +/- 20 years history expectations have evolved dramatically. In the early days the cochlear implant was considered as a lip reading aid, whereas nowadays user expectations include good hearing in noise and enjoyment of music. Faithful encoding of these complex auditory scenes constitutes a major challenge in terms of sound processing and understanding the inter-subject variation in the interface between electrode and auditory nerve. In the talk we present the linear algebra and optimization techniques we use to address these challenges.
18-03-2009
Luc Beirens (Federal Computer Crime Unit)
Cyber crime threats on e-world: How to prepare for and how to survive Cyber crime
Luc Beirens is the department head of the Federal Computer Crime Unit
(FCCU).
29-04-2009
Dr. Ir. Chris Tampère & Dr. Ir. Francesco Viti (Katholieke Universiteit Leuven)
Mathematical applications in traffic and transportation: traffic flow theory, queuing models and network equilibria
A traffic network is a physical network of multi-commodity traffic
flows. Mathematical models are used to describe and predict various
processes in such networks, ranging from models for individual driver
and vehicle behaviour over models for the aggregate behaviour of traffic
on the arcs and nodes, to models describing the overall equilibrium or
transient conditions in the network as a whole. Unlike physical systems
or fully automated environments, human behaviour plays a dominant role
in traffic networks. As a consequence of the heterogeneity and
unpredictability of human behaviour, mathematical models for traffic
only roughly capture the behaviour of the real system and can never
reach the level of accuracy of for instance weather prediction models.
Just like in economics, the aim of the mathematical model is to capture
the complex interaction of multiple actors and influences and to
understand properties and behaviour of the traffic system. However, the
reliability of predictions using these models largely depends on the
data underlying it, on the way these data were processed and on careful
interpretation of the model's output, rather than on mathematical
high-tech used in the theoretical foundation or numerical evaluation of
the model.
In this presentation, we introduce some submodels that are used in the
analysis of traffic networks. First we discuss macroscopic and
mesoscopic approaches for describing traffic flows on arcs of the
network. Then we discuss a queuing theoretical approach for modelling
the dynamics of uncertainty of queue formation and spillback at
intersections. Finally, we briefly highlight mathematical equilibrium
issues in (dynamic) traffic networks.
20-05-2009
Prof. Dr. Walter Daelemans (Universiteit Antwerpen)
Progress in Natural Language Understanding
Full automatic natural language understanding (NLU) is an old ideal of Artificial Intelligence. Achieving this goal would make possible high quality solutions in Machine Translation, conversation with information systems in natural language, knowledge extraction from text etc. It became clear quickly that NLU is an "AI-complete" problem that involves common sense reasoning, world knowledge, and intricate inference processes. In other words, deep semantic analysis is needed. However, this can only be achieved for very small domains (toy problems) with a lot of work (explicitly entering lots of knowledge into the system). Computational Linguistics forgot about deep semantic analysis and NLU for a while, and switched to methods based on automatic statistical and machine learning-based analysis of corpora (text mining) that are robust and efficient, but lead to only superficial understanding with limited accuracy. Recent developments in computational linguistics and the availability of large amounts of encyclopaedic data on the web (Wikipedia) may make it possible to extract world knowledge and inference processes automatically. This would make possible a new generation of language understanding systems that combine the robustness and efficiency of text mining with the deep semantic analysis and accuracy of the old AI NLU systems. I will illustrate what is currently possible and speculate on what could be done with even more wikipedia-like data.
Welcome on the web site of the WIS, a seminar at the department of mathematics and computer science of the University of Antwerp. The seminar is organized by a nonconstant set of graduate students.
The seminar is an activity of the Departemental Doctoral Study Programme.
Here you can find a road map to the Middelheimcampus, building G. You can use this site to determine your itinerary.
Organization
