translated by Google

Seminar program for 2008/2009

Autumn 2008

18. 9. 2008
Introductory Seminar Seminar
Information on Seminar Concept in the Autumn Semester.
Agenda of the seminar.
25. 9. 2008
Michal Hejč: Data Quality Model
The presentation begins with a brief overview of the research in the field and with the basic terms definitions, heading to a new model definition. It introduces a prototype of the model and also a model of data value. The use of the model and its role in the process of data quality measurement is also discussed. The concept is illustrated by the case study of South Moravian waste management data and the approach of the Ministry of Environment of the Czech Republic. Conclusion suggests the future use of the model for data from the web community.
2. 10. 2007
Václav Němčík: TBA
At present, anaphora resolution is one of the greatest challenges in the field of natural language understanding. Although anaphora plays an important role in human communication and is in its essence an interdisciplinary issue, it is not widely familiar to computer scientists. Therefore, this talk provides a brief insight into the relevant linguistic background, and an overview of the various falvors of anaphora and their computational aspects. Finally, it mentions state-of-the-art methods for AR and sketches research aims on the way to a badly-needed anaphora resolution system for Czech
9. 10. 2008
Jan Vlach: Gaussian Quantum Marginal Problem
If we have some composite system in a given quantum state, we might ask if some reduced states are compatible with the state of the whole system in terms of its spectrum. This question is very important because it is common to investigate large quantum systems by local unitary operations. Gaussian version of the quantum marginal problem deals with Gaussian states which can be described only by their first and second moments of canonical coordinates. For 3 modes this problem is solved but I will try to show that it is possible to characterize these reductions generally.
16. 10. 2008
Tomáš Hnilica: Visualization of large FEM models
Today, numerical simulation is an indispensable part of the computer aided product development chain and FEM simulations are widely used. The amount of data produced by a finite element of calculation places a particular challenge on scientific visualization. Several optimization techniques and visualization approaches are known. The aim of my work is to develop a fast visualization engine for very large FEM models and simulation results that will be processed on commodity hardware. In this presentation, the FEM model structure and simulation results will be given. Survey to known optimization approaches and their feasibility for FEM will be discussed.
23. 10. 2008
Jan Pomikalek: Even larger web corpora
Corpus text has a valuable resource for many fields in computational linguistics. As a result of the Zipf law, many events in natural languages ​​occur rarely and we often do not have enough data to be able to study these events. Even though there is an enormous amount of text available on the web for some languages, the size of web corpora created to date has not yet exceeded 3 billion words. This presentation will describe a step-by-step procedure for creating a web corpus of English texts with a target size of 20 billion words. Related problems will be described with a focus on detecting near-duplicate documents in text-based collections, and will be presented with an original, effective solution to the problem.
30. 10. 2008
Jiří Filipovič General Purpose Computing on Graphics Hardware
The current fastest GPUs outperform today's CPUs by about an order of magnitude in floating point arithmetic and memory bandwidth. In addition, CUDA (Compute Unified Device Architecture) enables developers to write algorithms for GPUs directly in C-like language instead of reformulating them as graphics problems. These two facts make GPUs more attractive for general purpose computing. In this talk, the architecture and programmability of GPUs based on nVidia G80 will be presented. The examples of current successful GPU implementations outperforming the CPUs will be introduced and basic optimization problems specific to GPUs will be discussed. I will conclude with the outline of my research in this area.
6. 11. 2008
Ondřej Daněk: Graph Cut Based Image Segmentation
Image segmentation is one of the fundamental tasks in image analysis for which many methods have been proposed. In recent years well-founded methods based on combinatorial graph cut algorithms have emerged and have been successfully applied to a wide range of problems in vision and graphics. In this talk, the basic idea behind the graph based image segmentation will be presented along with its main advantages and disadvantages and some preliminary results demonstrating the application of graph-based methods in a fully automatic segmentation of cell nucleus clusters will be shown and discussed.
13. 11. 2008
Tomáš Čapek: State-of-the-art of semantic networks and lexicons
20. 11. 2008
Vojtech Krmicek: High-Speed ​​Network Traffic Acquisition and Preprocessing
The talk will present a design of a high-speed network traffic subsystem suitable for agent-based intrusion detection systems. This is a hardware-accelerated probe, which provides real-time network traffic statistics. Network traffic is stored in collector servers and preprocessed data is then sent to detection agents using heterogeneous anomaly detection methods. Presented system is designed to improve the performance of agent-based intrusion detection systems and enable them to efficiently identify malicious traffic. The main contribution of the presented system is its ability to aggregate real-time network-wide statistics from geographically dispersed probes. Traffic acquisition system is designed for deployment on high-speed backbone networks.
27. 11. 2008
Zdeněk Vrbka: 'Why are the traditional testing approaches not enough in service systems?
At present, the world economy shifts from product paradigm to service paradigm. The service systems deal with complex problems that can not be solved by using product approach. Service systems involve people, organizations, information and technology and their failure can have large, often catastrophic consequences. Therefore, it is important to test these systems in a proper way. The presentation illustrates the reasons why traditional testing approaches are not sufficient in the service system. It also draws attention to the need to study the problem of service systems testing.
4. 12. 2008
Martin Maška: A Two-Phase Cell Nucleus Segmentation Using Topology Preserving Level Set Method
An accurate localization of the cell nucleus boundary is inevitable for any further quantitative analysis of proteins, genes, chromosomes and other subnuclear structures within the cell nucleus. In this talk, we present a novel approach to the cell nucleus segmentation in fluorescence microscope images exploiting the level set framework. The proposed method works in two phases. In the first phase, the foreground image is separated from the background to obtain a binary mask of individual cell nuclei as well as their clusters. The second phase is focused on the boundary detection of each cell nucleus within the previously identified clusters.
11. 12. 2008
Jan Sedmidubský: A Self-organized System for Content-Based Search in Multimedia
We propose a self-organized system for content-based search in multimedia data. In particular, we build a semantic overlay over an existing peer-to-peer network. The self-organization of the overlay is obtained by using the social-network paradigm. The connections between peers are formed on the basis of a query-answer principle. The knowledge about answers to previous queries is exploited to route queries efficiently. At the same time, a randomized mechanism is used to explore new and unvisited parts of the network. In this way, a self-adaptable and robust system is built. Furthermore, the metric-space data model is used to achieve extensibility. The proposed concepts are verified on a network of 2,000 peers and indexing 10 million images.
18. 12. 2008
Poster Session

Spring 2009

19. 2. 2009
Introductory Seminar of the Spring Semester
Information on the seminar concept in the spring semester. Agenda of the seminar. Discussion.
26. 2. 2009
Nikola Benes: Partial Order Reduction for State / Event LTL
Software systems assembled from a large number of autonomous components become an interesting target for formal verification due to the issue of correct interplay in component interaction. State / event LTL incorporates both states and events to express important properties of component-based software systems. The main contribution of this work is a partial order reduction technique for verifying LTL properties. The core of the partial order reduction is a novel notion of stuttering equivalence which we call state / event stuttering equivalence. The positive attribute of the equivalence is that it can be resolved with existing methods for partial order reduction. State / event LTL properties are generally not preserved under state / event stuttering equivalence. To this end, we define a new logic called weak state / event LTL, which is invariant under the new equivalence.
5. 3. 2009
Jan Vykopal: Network-based Dictionary Attack Detection
This paper describes the novel network-based approach to a dictionary attack detection with the ability to recognize successful attack. We analyzed the SSH break-in attempts at a flow level and determined a dictionary attack pattern. This pattern has been verified and compared to common SSH traffic to prevent false positives. The SSH dictionary attack pat- tern was implemented using decision tree technique. The evaluation was carried out in a large high-speed university network with promising results.
12. 3. 2009
Jan Vlach: Quantum marginal problem
A very interesting question is what multipartite Gaussian states can be prepared. This presentation deals with the relationship between symplectic spectra of n-mode quantum systems and symplectic spectra of their single-mode subsystems. It shows results for two cases when both of the subsystems are in pure or in mixed quantum state. This talk provides a brief overview of the results obtained and presents new methods and ideas used to look for valid conditions for more general reductions to subsystems consisting of k modes.
19. 3. 2009
Jiří Materna: Czech Verbs in FrameNet Semantics
In the natural language processing field, there is a trend to build large electronic lexical databases based on semantic information. These resources are widely used in several applications such as information retrieval, machine translation and even disambiguation tasks on all levels. This work presents a method of automatic verbs and their valencies in VerbaLex database to entries in Berkeley FrameNet. While a complete manual work can take a long time, this automatic approach only requires a bit of human effort to reach sufficient results. By linking VerbaLex to FrameNet, we are able to find a nontrivial subset of interlingual FrameNet frames (including their frame-to-frame relationships), which could be used as a base for building FrameNet in Czech.
26. 3. 2009
Milan Ceska: Local Quantitative LTL Model Checking
Quantitative analysis of probabilistic systems has been studied mainly from the global model checking point of view. In global model-checking, the goal of verification is to determine the probability of satisfying a given property for all available states in the state space of the system under investigation. On the other hand, in the local model checking approach, the probability of satisfaction is calculated only for the set of initial states. In theory, it is possible to solve the local model checking problem using the global model checking approach. However, the global model checking procedure can be significantly outperformed by a dedicated local model checking one. In this paper we present several specific local model checking techniques that reduce the runtime needed from days to minutes.
2. 4. 2009
Vojtěch Kovář: Automatic Processing of Czech Syntax
The goal of the natural language syntactic analysis is to reveal the surface structure of the input text, or input sentence, respectively. It can be viewed as a cornerstone of any complex natural language processing tasks ranging from intelligent searching in text to question answering systems and complex information analysis of input text. In the presentation, we give an overview of the current approaches to the automatic syntactic analysis of the Czech language. We will describe formalisms used for encoding syntactic information, available annotated data and measurement techniques as well as selected parsers and parsing algorithms. Also, most remarkable problems in this field will be discussed and their possible solutions outlined.
9. 4. 2009
Roman Žilka: Electronic Micropayment Schemes
When you are paying your health insurance, a conference entrance fee or sending money to your employees, it's perfectly reasonable to use ordinary bank transfers. These payments are infrequent enough and high enough to justify the complex transfer operation. However, it would be an overkill to use classic transfers to purchase separate web articles worth 1 CZK each as you browse an e-zine site, or separate video clips as you browse a vlogging site. These frequent and tiny payments need special handling. Although it is not very widespread around the Net these days, some ideas for the future, such as on-demand computing, call for a versatile and simple scheme to allow for these "micro-payments", as the self-explanatory term goes. Two such schemes will be discussed.
16. 4. 2009
Ziad Salem: Current Research Projects at the Electrical and Electronic Engineering Faculty, Computer Engineering Department of Aleppo University.
The talk will begin by giving some information about Syria, then Aleppo University, then master course at the Computer Engineering Department, Electrical and Electronic Engineering Faculty. Then I will give a brief description of the project which I supervise there.
23. 4. 2009
Vojtech Krmicek: NetFlow Based Monitoring in the FEDERICA Project
An important part of today's and future high-speed networks is reliable and detailed traffic monitoring. Especially important is the traffic monitoring in research networks like FEDERICA. These networks consist of physical layer and virtual layers above. Researchers need to know detailed statistics about traffic in virtual layers in virtual networks, which are complicated to obtain by standard monitoring tools. This presentation describes a concept of virtual network monitoring based on extended NetFlow. The reconstruction of the virtual network topology as well as the use of the fine grain flows method will be presented.
30. 4. 2009
J. Plhák: Dialogue-Based Processing of Web Presentation
In this presentation, we describe the basic methods and technologies used in the BrowserWebGen system prototype, which allows blind users to develop their own web presentation using a dialog. This approach benefits especially from the web browser and screen reader software. The basic principles of the BWG system are discussed as well as the comparison of this system and the VoiceXML solution. As an illustration, we provide an example showing how the blind can create the web page.
7. 5. 2009
Martin Šmérek: I / O-Effective Binary Decision Diagram Manipulation
Model checking is a popular approach for formal verification of reactive systems. However, the use of this method is limited by the so-called state space explosion. One way to cope with this problem is to represent the model and the state space symbolically by using Binary Decision Diagrams (BDDs). Unfortunately, during the computation the BDD can become too large to fit into the available main memory and it becomes essential to minimize the number of I / O operations. We present an extension of existing algorithms for BDD manipulation and propose a new I / O-efficient algorithm for calculating existential quantification.
7. 5. 2009
Ivan Fialik: Cryptographic Applications of Pseudo-Telepathy Games
Communication complexity is an area of ​​classical computer science which studies how much communication is necessary to solve various distributed computational problems. Quantum information processing can be used to reduce the amount of communication required to carry out some distributed problems. We speak of pseudo-telepathy when it is able to completely eliminate the need for communication. After introducing a general model for pseudo-telepathy games and some necessary cryptographic definitions, we describe a simple user identification protocol based on playing some pseudo-telepathic game by the parties.