Program kolokvií pro semestr Podzim 2004

5. 10. 2004

Dr. Martin Rajman, Swiss Federal Institute of Technology, Laussane

Thematic Annotation: extracting concepts out of documents

Abstract: Topical annotation - i.e. the annotation of document segments with tags identifying the topics discussed in the segments - might be used in many tasks, and, in particular, to enhance information retrieval. Unlike standard approaches to topic annotation, the presented technique does not centrally rely on some sort of - possibly statistical - keyword extraction. Instead, the proposed annotation algorithm uses a large scale semantic database - the EDR Electronic Dictionary - that provides a concept hierarchy based on hyponym and hypernym relations.This concept hierarchy is used to generate a synthetic representation of the document by aggregating the words present in topically homogeneous document segments into a set of concepts best preserving the document's content. The proposed approach for topic extraction was experimentally tested and evaluated on a database of 238 documents corresponding to bibliographic descriptions extracted from the INSPEC database. A novel evaluation metric was designed to take into account the fact that the topics associated with the INSPEC descriptions - taken as the golden truth for the evaluation - were not produced based on the EDR dictionary, and therefore needed to be approximated by the available EDR entries.

12. 10. 2004

Dr. Ing. Otto Fučík, FIT, VUT Brno

HW/SW Co-design

Abstract: Příspěvek ze zabývá metodikou návrhu komplexních číslicových systémů, ve kterých jsou jednotlivé subsystémy implementovány jak v software, tak i v hardware. Cílem souběžného návrhu hardware a software (Hardware/Software Codesign) je maximalizace poměru výkon/cena při realizaci dané aplikace - tedy nalezení optimálního členění úlohy na "pomalou, ale levnou" softwarovou a "rychlou, ale drahou" hardwarovou implementaci aplikace a zajištění efektivní interakce mezi jednotlivými částmi. Bude též prezentováno použití uvedeného přístupu při realizaci praktických aplikací - směrovače paketů IPv6 a systému pro zpracování obrazu v reálném čase.

19. 10. 2004

Prof. Philippe Jorrand, Leibniz Laboratory, Grenoble, France

Quantum: from Physics to Programming

Abstract: Information is physical: the laws which govern its encoding, processing and communication are bound by those of its unavoidably physical incarnation. In today's informatics, information obeys the rules of classical Newtonian physics: this statement holds all the way from commercial computers down to (up to?) Turing machines and lambda-calculus.
The driving force of research in quantum computation is that of looking for the consequences of having information encoding, processing and communication based upon another kind of physics, namely quantum physics, i.e. the ultimate knowledge that we have, today, of the foreign world of elementary particles, as described by quantum mechanics. A rapid survey of the principles of quantum information processing and communication shows that algorithms and protocols making use of quantum resources always comprise both quantum and classical components, assembled together so that they communicate and cooperate. Indeed, for the results of quantum information processes to be useful in any way, there is an intrinsic necessity for organized cooperation and communication between quantum and classical devices, controlled by the classical side. This simple, obvious fact has profound consequences on a number of central issues in computer sciences, like machine architectures, language design, semantics, type theories, abstract models of computation, etc., as soon as quantum primitives and resources are accessed by information processes. These issues cannot be ignored if the aim is to reach a satisfactory understanding of what is the meaning, what is the power, what are the limitations of computing and communicating with the help of quantum resources.
After introducing these bases, this seminar will present a recent contribution in that direction, concerning the design of a quantum communicating process calculus, where quantum and classical parts of information processes can be described and assembled consistently, while making sure that the peculiar properties of the quantum world are guaranteed by the semantics. Communication protocols making use of quantum resources will be given as examples, illustrating this approach to the rigorous description of quantum+classical information processing.

26. 10. 2004

Doc. Vítězslav Švalbach et al., FI MU, Brno

Ateliér grafického designu a multimédií a jeho výsledky a místo ve struktuře (výuky) Fakulty informatiky

Abstract: Stručná historie (místo úvodu). Současný stav vyučovaných předmětů. Integrace výuky do struktury Fakulty informatiky (specializace, bakalářské a diplomové práce). Profil absolventa. Personální obsazení. Prezentace vyučovaných předmětů. Bezejmenná galerie. Doprovodné akce. Aktivity pedagogů. Aktivity studentů. Připravované akce. Vize.

2. 11. 2004

JUDr. František Cvrček, CSc., Ústav státu a práva AV ČR v Praze

Právní informační systémy a právní jazyk

Abstract: Cílem přednášky je seznámit posluchače s výsledky aktuálních výzkumů prováděných Laboratoří právní informatiky (dále LPI) Ústavu státu a práva Akademie věd ČR. LPI vytváří právní databáze od roku 1985. Tyto databáze obsahující právní předpisy od roku 1790, judikaturu od roku 1918 a vybranou literaturu od roku 1990 tvoří východisko prováděných analýz. LPI se věnuje pravidelně aktualizované deskripci právního řádu, tvorbě právních informačních systémů a analýze právního jazyka na úrovní textů právních dokumentů. Výsledky práce LPI vedly k zboření některých mýtů panujících v právní vědě, např. existence tzv. legislativní inflace, konzistence právního systému, jednoznačnost právní terminologie, úloha racionality v právním usuzování, explikační síla tradičních právních klasifikací, apod. Od roku 1990 pracovníci LPI úspěšně vyřešili 10 projektů GA AV ČR, GAČR, podílejí se na mezinárodních projektech podporovaných CRN v Itálii a Evropskou komisí (dále EK). Jsou zakladateli oboru Právní informatiky, které se od roku 1993 vyučuje na všech právnických fakultách v ČR.

9. 11. 2004

Prof. Eduard Groeller, TU Wien

Importance-Driven Image Generation in Scientific Visualization

Abstract: Visualization is the use of computer-supported, interactive, visual representations of data to amplify cognition. After a quick introduction to scientific visualization several previous projects in medical visualization at the Vienna University of Technology will be surveyed.
Importance-driven volume rendering as a novel technique for automatic focus and context display of volumetric data is presented in detail. The technique is a generalization of cut-away views, which - depending on the viewpoint - remove or suppress less important parts of a scene to reveal more important underlying information. For each object several representations, i.e., levels of sparseness, are specified. The display of an individual object may incorporate different levels of sparseness. The goal is to emphasize important structures and to maximize the information content in the final image.
The last part of the talk discusses the VesselGlyph as an importance-driven display of CT-Angiography data. The VesselGlyph covers several regions where different rendering methods are used. The region type, the used visualization method and the region parameters depend on the distance from the blood vessel centerline and on viewing parameters as well. Typically direct volume rendering (DVR) and curved planar reformation (CPR) is combined within a single image.
Further information on the research projects is available at http://www.cg.tuwien.ac.at/research/vis/

16. 11. 2004

Dr. Martin Leucker, TU Munich

Message-Passing Automata are expressively equivalent to EMSO Logic

Abstract: We study the expressiveness of finite message-passing automata with a priori unbounded channels and show them to capture exactly the class of MSC languages that are definable in existential monadic second-order logic interpreted over MSCs. Moreover, we prove the monadic quantifier-alternation hierarchy over MSCs to be infinite and conclude that the class of MSC languages accepted by message-passing automata is not closed under complement. Furthermore, we show that satisfiability for (existential) monadic seconder-order logic over MSCs is undecidable.

23. 11. 2004

Dr. Joern Mueller-Quade, PhD., Universitaet Karlsruhe

What does cryptographic security mean?

Abstract: The question if a protocol is secure strongly depends on how security is defined. Starting with an intuitive and very informal definition of security we present several problems and drawbacks of an ad hoc approach to cryptographic security.
Next we give an overview of a more modern definition of security which is so strict that it even guarantees the property of security being preserved under composition. With such a strict notion of security a modular design of larger protocols becomes possible.
Unfortunaltely even the new definitions of security do not cover all aspects of security and we give a critical view on the models underlying the security definition. We will in particular show that the notion of security depends on the quality of the underlying network (modelled by the influence the adversary has on the message delivery) and that, paradoxically, some cryptographic tasks cannot be realized if the quality of the network is too good.

30. 11. 2004

Doc. Petr Hliněný, PhD., TU Ostrava

On Matroids, MSO Logic, and Computation

Abstract: Matroids present a strong generalization of graphs, and some recent deep results of structural matroid theory have found (somehow surprising) applications back in graph theory. Likewise we may mention the approximation algorithm for graph clique-width by Oum and Seymour, and the related WQO results. We give a brief introduction to structural matroid theory, and show its close relation to the Graph minor project of Robertson and Seymour. We define matroid branch-width (and tree-width), showing their basic properties. We also discuss matroidal analogues to the results of Courcelle and others on computability of MSO-definable graph properties on bounded tree-width graphs.

7. 12. 2004

Ing. Jan Žižka, CSc., FI MU, Brno

Discovering Complex Patterns by Machine Learning

Abstract: Humans are generally able to solve difficult problems just by learning from specific and generalized examples that provide knowledge -- especially if there is no suitable, analytical, and exact way to define and complete complex tasks. Ideally, machines should be able to learn similarly as humans do. Chess is an excellent and established model for testing artificial intelligence tasks.
This talk deals with a particular pattern recognition by machine learning. The patterns are specific chess positions where the white player can sacrifice the bishop (placed on the diagonal b1-h7) on the square h7: Bxh7+. A computer learns if the sacrifice leads to a won or lost game, i.e., a classification task based on results of examples of real games. As a learning algorithm, the entropy-based decision tree is used. The algorithm does not employ any chess rules or calculations of positions, it simply learns from a selected set of 219 positive and 231 negative training examples with eight different representations of real positions played originally by expert human players in the years 1601-1999. The correct recognition ranges from 87% to 93% depending on a particular representation of game positions.
The object is to demonstrate that machines can learn to quickly recognize typical situations and then to decide which possibility should be analyzed preferentially.

14. 12. 2004

Dr. Josef Pojsl, Trusted Network Solutions

Trendy ve vývoji firewall

Abstract: Dvě základní technologie pro firewall jakožto zařízení oddělující sítě s různou úrovní důvěryhodnosti jsou stavové IP filtry a aplikační proxy brány. Jejich principy se již několik let prakticky nemění.
Problém obou spočívá v tom, že jsou ve své základní podobě často velmi málo účinné proti moderním hrozbám. Útoky se skrývají ve zdánlivě legitimním provozu, takže např. rozeznat URL, které způsobí napadení klienta červem, jež se začne šířit po celé interní síti a prakticky ji zahltí, je obecně velmi obtížný problém.
Firewally se proto začínají kombinovat s dalšími technologickými nástroji, aby mohly efektivněji plnit svou roli. V přednášce představíme současné možnosti integrace firewallu s antivirovými programy, detektory průniku a dalšími nástroji zkoumajícími obsah komunikace a zamyslíme se nad budoucím potenciálem tohoto přístupu.