translated by Google

Seminar program for 2009/2010

Autumn 2009

25. 9. 2009
Introductory Seminar Seminar
Program:
Information on Seminar Concept in the Autumn Semester.
Agenda of the seminar.
Discussion.
1.10. 2009
Marek Grac: Building language resources with community
Abstract:
The complexity of natural languages ​​is still beyond the capacity of computers. Attempts to its formalization have led to the obtaining of language resources in various layers (eg word forms or dictionaries). Their coverage and precision are tested on existing text corpora. There are situations when it is appropriate to use men power to handle specific cases. These tasks are usually very exhaustive and cross-checking is very important. This presentation will describe our attempt to work with community on building morphological database. Related issues as motivation and targeted selection are covered as well.
8.10. 2009
Radek Krejčí: Netopeer - Secure Remote Configuration System
Abstract:
TBA
15.10. 2009
Fedor Tirss: Assisted photography for the blind
Abstract:
Presentation briefly introduces a concept of assisted photography for blind and visual impaired users. We also describe the architecture of GATE (Graphics Accessible to Everyone) and WEBGEN in combination with PHOTON. Today's smartphones are equipped with the localization service (GPS) and the digital compass. In conjunction with the camera and internet access, it allows the blind to take their own photos. This kind of photos can be easily annotated, tagged and posted on the website. As an illustration, we provide an example showing how the blind can investigate (by sound or dialog) the "smart" photos on the web. This approach is not limited to blind, but also allows the commercial use of a whole concept.
22.10. 2009
Antonín Pavelka: Predicting effect of substitution on protein function
Abstract:
A single amino acid substitution can alter the protein structure so dramatically that the protein is not able to perform its function anymore. The ability to predict if this happens is important for both protein engineering and biomedical research. We use the receiver operating characteristic and the accuracy-coverage curve to show that the confidence weighted consensus of current methods leads to increased predictive performance. We also suggest plans for further improvement and deployment in protein engineering.
29.10. 2009
Vojtěch Bystrý: Segmentation of protein sequences using hidden Markov models
Abstract:
TBA
5.11. 2009
Vít Baisa: Machine Translation: Methods of Evaluation
Abstract:
If we want to compare different machine translation systems (tools, programs) we need to have our metrics available at our disposal. There are several methods commonly used to measure and quality translation and they will be presented in the talk. It will also be pointed out to their limitations and shortcomings and will show several ways of dealing with these limitations.
12.11. 2009
Zdeněk Vašíček: Evolutionary design of image filters and its hardware acceleration
Abstract:
Although the design of linear filters is a straightforward process, the design of robust nonlinear filters represents a non-trivial task and thus the median filter remains (since its invention) the most used nonlinear filter. We will show that the evolutionary design (evolutionary algorithms) approach can be successful in solving this task. In this presentation, a method will be introduced to design robust nonlinear image filters using evolutionary techniques. It will be shown that our approach produces solutions that show better quality as well as lower implementation cost compared to human-designed median-based filters. Since evolutionary design is generally a time-consuming process, an FPGA-based platform has been developed to accelerate this task. This platform will be discussed in the second part of the presentation. Since the modern XILINX FPGAs contain embedded PowerPC processors, the search algorithm is implemented using PowerPC while the candidate solution is evaluated using a deeply pipelined architecture. It will be shown that a significant acceleration of evolution is obtained when comparing to a highly optimized software implementation even if evolutionary platform runs at 100MHz. In the last part, the results will be discussed.
19.11. 2009
Jiří Kolář: Self-optimizing business processes
Abstract:
TBA
Petra Budíková: Response quality refinement in content-based information retrieval
Abstract:
Since traditional databases are not suitable for the efficient management of multimedia data, a new content-based approach and similarity research has been intensively researched in recent years. However, the concept of similarity is tricky - it will be shown that no universal measure of similarity can be defined. Therefore, the search system has to cooperate with the user. We will briefly present the concept of relevance feedback (RF) and discuss the difficulties that need to be overcome to allow successful utilization of RF in similarity searching. The main focus will be on the efficiency issues - we will analyze the cost of RF for a real-world search system and propose several methods of reducing costs.
26.11. 2009
Lukáš Holík (FIT BUT): Nondeterministic tree automate for abstract regular model checking
Abstract:
Abstract regular tree model checking (ARTMC) is a generic framework for verifying infinite state systems with tree-shaped configurations (such as pointer manipulating programs, parametrated protocols, ...). Possible infinite sets of states are represented as tree automata. Nondeterministic tree automata (NTA) is much more compact then deterministic ones, however, to use NTAs within ARTMC, one has to develop techniques for dealing with NTAs effectively without determining them. Such techniques will be presented, specifically, algorithms for reducing size and for checking the language inclusion of NTAs.
3.12. 2009
Petr Jarušek: How to Predict the Difficulty of Sokoban Puzzle?
Abstract:
Sokoban is a well-known Japanese game proven to be a P-SPACE complete problem. Although the game is hard to solve for computers, people are often able to solve it very guickly. What causes difficulty of Sokoban puzzle for humans? How can we predict the difficulty of other puzzles? What is it all good for? We will focus more on our human-like computational model and possible applications of our research. (see also http://en.wikipedia.org/wiki/Sokoban http://organizatori.cz/puzzle/)
10.12. 2009
Vojtech Kovar: Syntactic Analysis as Pattern Matching
Abstract:
TBA
17.12. 2009
Poster Session
Abstract:
Presentations will be presented in the autumn semester.

Spring 2010

19. 2. 2010
Introductory Seminar of the Spring Semester
Information on the seminar concept in the spring semester. Agenda of the seminar. Discussion.
4.3. 2010
Marek Trtík:
Abstract:
TBA
11.3. 2010
Jan Vlach: Quantum marginal problem
Abstract:
The presentation will deal with the compatibility of quantum states of the composite system and given quantum states of its subsystems. We will show known results for the case of finite-dimensional Hilbert spaces. For continuous quantum variables there is no known general solution. However, for certain classes of Gaussian states this problem can be significantly simplified.
18.3. 2010
Zuzana Nevěřilová:
Abstract:
TBA
25.3. 2010
Tomáš Čapek: Construction and Maintenance of Semantic Networks
Abstract:
A general-purpose semantic network is a language resource for the given language, alternative to traditional dictionaries. It consists of semantic units that are connected by semantic relations, thus creating a graph-like structure or a network. In this presentation we will discuss how semantic networks are constructed and how to maintain their data in a semi-automatic way.
1.4. 2010
Michael Tsikerdekis: Social networks & online group phenomena
Abstract:
The presentation is divided into two sections. In the first one I'm going to talk about social networks and explain what they are, how they work, and give some examples of different categories. Then I'm going to talk about issues related to social networks and specifically about design flaws that could lead to economic problems as well as ethical problems and give examples. In the second section I am going to talk about two specific phenomena that occur in online groups and thus in social networks. I'm going to point out the causes of this phenomenon and discuss that software engineers should use this knowledge to avoid design flaws and predict online group behavior to meet the desired end result.
8.4. 2010
Jana Tůmová: Optimal Path Planning under Temporal Logic Constraints
Abstract:
We present a method for automatically generating optimal robot trajectories that meet high level mission specifications. The motion of the robot in the environment is modeled as a general transition system, enhanced with weighted transitions. The mission is defined by a class of temporal logic formulas. For each environment model and for every formula in class, our method calculates a robot trajectory that minimizes a certain cost function. The problem is motivated by applications in robotic monitoring and data gathering.
15.4. 2010
Rambousek: TBA
Abstract:
TBA
22.4. 2010
Pavel Karas: TBA
Abstract:
TBA
29.4. 2010
Pavel Černohorský: TBA
Abstract:
TBA
6.5. 2010
Ondřej Strnad: TBA
Abstract:
TBA
13.5. 2010
Martin Drašar: TBA
Abstract:
TBA
Vilém Šustr: Computer Aided Calligraphy in Haptic Virtual Environment
Abstract:
TBA
20.5. 2010
Poster Session
Abstract:
Presentations will be made in the spring semester.