Programme of colloquia with abstracts for the Autumn 2005 semester

20 September 2005

Prof. Jana Kosecka, George Mason University, Washington, USA

Vision-Based Localisation in Urban and Office Environments

Abstract: The ability to localise in an unknown environment is essential for basic navigation tasks carried out by both people and robots. I will describe two localisation approaches for indoor and outdoor environments using images captured by inexpensive cameras. Localisation is achieved by means of location recognition followed by the computation of the relative pose of the query view with respect to known landmarks.

In the first approach, the topological model of the office environment is acquired during the exploration phase and comprises locations and the neighbourhood relationships between them. Given a new query view, its features are matched against those of the model views and integrated probabilistically, yielding an estimate of the most likely location. In indoor environments, ambiguities arising from self-similarity and dynamic changes in the environment are resolved by exploiting spatial relationships between locations captured by a Hidden Markov Model.

In the second system, the model comprises a database of buildings from an extended urban area. We propose an efficient image descriptor for the rapid retrieval of candidate matches and robust feature-based matching for the final pose computation.

27 September 2005

Prof. Václav Přenosil, CSc., FIMU

Simulation and training technologies in the improvement of decision-making processes for management staff

Abstract: The application of information technology makes the process of developing the skills of managers and operators in the management of resources and processes highly flexible. Computer technology is capable of creating conditions for practising decision-making and control processes, even in situations which, whilst likely to occur in reality, are very difficult to replicate. The development of information technology offers unprecedented opportunities for implementing constructive, virtual and live simulations for the training of both senior managers and process operators and technical equipment operators.

Modelling and simulation (M&S) methods and tools are currently capable of providing immediately available, flexible and cost-effective means of significantly improving and expanding activities in the areas of operational management, planning, training, exercises, operational safety, research, technological development and technology procurement.

For a number of activities (air traffic control, energy and communications systems, environmentally hazardous technological processes), this objective is supported by cooperative efforts on an international scale. These efforts will expand the scope for interoperability and the application of M&S tools to other areas of social process management, where the preparation and training of management staff has not yet been routinely carried out.

‘Coalition Plans’ for modelling and simulation with international scope are being established, creating the conditions for a cooperative approach to the application of modern simulation techniques with a view to helping meet the needs of the alliances and their member countries.

4 October 2005

Doc. Ivan Kalaš, PhD., FMFI UK Bratislava

Software tools supporting the development of cognitive processes

Abstract:

• Background – Piaget’s constructivism and Papert’s constructionism. Papert’s idea that computers also belong to children
• The computer in the cognitive process – what role does it play, and what role should it play? How does it contribute to more effective and engaging learning
• IT and digitalisation, the training of future teachers, and general and subject-specific information literacy
• new software platforms to support school computer science and computerisation
• what new features the Comenius Logo environment, animated images
• what’s new in the Imagine environment, programmable images, objects, processes and events
• Modern programming for children
• Imagine as a tool for developing educational software
• The CoLabs project: on international communication and collaboration amongst children. Colaboratory is an attempt to realise Kay’s Dynabook
• CoLabs and Kaleidoscope: a new, tangible approach to mathematics – the Visual Fractions environment and its benefits
  • a library of visual interactive objects, building blocks for discovering concepts and relationships (constructionism),
  • the features of the Imagine platform that make this possible,
  • a new tool for educational research,
  • new forms of collaboration and communication between pupils.
• TEL: new technologies for new forms of learning

11 October 2005

Doc. Petr Jančar, CSc., VŠB – Technical University of Ostrava

Some aspects of formal verification, with a focus on bisimilarity

Abstract: The author first plans to briefly survey the history and current trends in (some areas of) formal verification of computer systems (theorem proving, model checking, behavioural equivalences, etc.) The talk will then focus on bisimulation equivalence on infinite-state systems, and a recent result by Jančar and Srba will be presented, namely the undecidability of bisimilarity on so-called Type -1 systems (with rules R → w). This shows that the strategies used by Senizergues and Stirling in the decidability proofs for pushdown graphs cannot be generalised to Type -1, and their respective open question is thus resolved.

18 October 2005

Dr Vlastimil Havran, PhD, Max Planck Institute for Informatics, Saarbrücken

Sorting and Searching in Photo-Realistic Image Synthesis

Abstract: We demonstrate the role of sorting and searching in high-quality image synthesis, which is utilised in many computer graphics algorithms. We show that classical information organisation and processing are inherent and constitute the most important aspect underlying the computation of synthetic images. In detail, we discuss how so-called final gathering in photon mapping – used, for example, to generate ‘special effects’ images in films – can be accelerated by reversing the entire search process at the expense of memory usage. The speed-up achieved for searching is by an order of magnitude.

25 October 2005

Carlos Ortiz de Solorzano, PhD, University of Navarre, Pamplona, Spain

Contribution of 3D image analysis to understanding normal mammary gland biology and breast cancer

Abstract: Cancer occurs as a series of transformational events that alter the architectural organisation of tissue structures as well as the genetic and epigenetic characteristics of normal tissue cells. A great deal of quantitative information about these changes can be obtained using automatic or semi-automatic 3D image analysis tools and computerised microscopy. In this talk, examples of these tools and their application to the study of the development of the normal mammary gland and breast cancer will be presented.

1 November 2005

Prof. František Plášil, DrSc., Faculty of Mathematics and Physics, Charles University, Prague

Formal methods and software components – a real-life experience

Abstract: The talk is based on real-world experience with the behavioural specification of a non-trivial component-based application. The experience shows that model checking of such a specification yields very long error traces (providing counterexamples) involving hundreds of states. Analysing and interpreting such an error trace to localise and debug the actual fault in the specification is a tedious task. We present two techniques designed to address the problem: state space visualisation and protocol annotation, and share our positive experience of applying them, in terms of making the debugging process more efficient.

8 November 2005

RNDr. Jozef Vyskoč, PhD., Bratislava

Privacy protection – an overview of the current state

Abstract: The issue of privacy protection has been closely linked to technological progress from the very beginning (initially with the development of photographic techniques). However, it is not merely a matter of the ability to collect and process large amounts of data and deriving new conclusions from it – through the use of modern tools and their activities in cyberspace, people leave digital traces, the existence of which they are not always aware of. Privacy protection is an endeavour to find a suitable compromise between the legitimate use of such (not only digital) traces and the requirements of the individuals concerned to have control over information that reveals certain characteristics, habits and so on.

This lecture will attempt to provide an overview of the current state of affairs in the field of privacy protection, both from the perspective of ‘theory’ (related concepts and attempts to define them, the law) and from that of practice (personal experience from security projects, the results of a survey carried out in early 2005 in Slovakia, and a comparison of the Czech and Slovak personal data protection laws).

15 November 2005

Prof. Jiří Wiedermann, Institute of Informatics, Czech Academy of Sciences, Prague

RNDr. Dana Pardubská, PhD., Comenius University, Bratislava

On the Power of Broadcasting in Mobile Computing

Abstract: A computational model reflecting fundamental computational aspects of wirelessly communicating mobile processors is presented. In essence, our model is a deterministic Turing machine capable of launching new processes, among which wireless communication via explicitly assigned channels must be programmed. We demonstrate that computations on such machines are polynomially time- and space-equivalent to the synchronised alternating Turing machines previously studied in the literature. This shows that nondeterminism can be completely eliminated from synchronised alternation at the cost of introducing programme-driven communication amongst the respective processors. We also study the log space and poly time hierarchies between LOGSPACE, PTIME and PSPACE induced by bounds on the number of channels used in parallel broadcasting.

22 November 2005

Prof. Petr Hájek, DrSc., Institute of Logic, Academy of Sciences, Prague

Mathematical Fuzzy Predicate Logic – Current Status

Abstract: What is logic, mathematical logic, mathematical fuzzy logic? The basic concepts and properties of mathematical fuzzy predicate logic will be explained as a multi-valued logic whose standard set of truth values is the real unit interval and whose semantics is based on the concept of continuous t-norms; the general semantics is defined by the concept of BL-algebras (all terms will be properly defined). The aim of the lecture is to outline the current state of research in this remarkable discipline of mathematical logic.

29 November 2005

Dr Christopher Kruegel, TU Vienna

Malicious code analysis: Detecting metamorphic worms

Abstract: Computer worms are malicious programmes that spread autonomously across communication networks by exploiting vulnerabilities affecting a large number of hosts. The intensity and frequency of worm outbreaks have increased dramatically in recent years, and each successful epidemic causes significant economic damage. Thus, there is a profound need for efficient countermeasures that can stop a worm in its early stages.

Current worm detection systems are typically based on syntactic signatures, which specify byte sequences that are characteristic of a particular worm instance. These systems rely on the common assumption that the worm code does not change during the infection process. Unfortunately, worms can be metamorphic. That is, the worm’s code can mutate whilst spreading across the network, thereby rendering detection using signatures ineffective.

In this presentation, I will discuss research challenges and our proposed solutions to the problem of identifying metamorphic worms in network streams. In particular, I will introduce novel techniques that are capable of characterising similarities between different mutations of a worm. These similarities are captured by abstract models that describe the behaviour and structure of malicious code. The key idea is that structural and semantic properties are more difficult to alter with each new generation of the worm. Consequently, our approach yields a more general and robust description of malicious code that is unaffected by syntactic changes in the binary image.

6 December 2005

JUDr. František Cvrček, CSc., Institute of State and Law, Czech Academy of Sciences

Legal language, analysis of legal texts and law-making

Abstract: The characteristics of legal language as a specialised language at the level of dictionaries and the structures of legal documents. Project results: 1. Quantitative description of the Czech legal system, 2. Legal terminology and 3. Contextual legal dictionary. Problems with the European LOIS project (an attempt to create a multilingual dictionary based on WORDNET technology). A comparison of standard law-making at the level of drafting bills with practice in the Czech Republic.

The lecture takes a critical look at legislative practice in the Czech Republic, based on the long-term findings of the Laboratory of Legal Informatics at the Institute of State and Law of the Czech Academy of Sciences. The above-mentioned projects were carried out using the LEGSYS database, which is a specialised legal database covering our legal system since 1918 in the areas of legislation and case law.

13 December 2005

Prof. Tomoyuki Yamakami, Japan Science and Technology Agency, Japan

Resource-Bounded Computation and Erroneous Probabilistic Computation

Abstract: We discuss two recent topics in computational complexity theory. First, we study a weak computation model where computation is limited to a single block of linear-size memory space (which also includes input information) and linear running time. Although this computation model is closely related to finite automata, it further introduces the notion of multi-valued partial functions. We can prove that there is no one-way function in our model. Second, we consider erroneous computation caused by two different types of errors associated with probabilistic computations. Of particular interest are so-called nearly bounded-error probabilistic computations. We show a relationship between such computation and non-uniform computation. Finally, we prove a fundamental theorem on nearly bounded-error probabilistic computations.