Programme of colloquia with abstracts for the Autumn 2008 semester

23 September 2008

Doc. RNDr. Petr Hliněný, Ph.D., FIMU, Brno

Approximation algorithms for the intersection number of graphs

Abstract: This talk will present new results for the important problem of computing the intersection number of graphs.

30 September 2008

RNDr. Eva Hladká, Ph.D., FIMU, Brno

Group communication

Abstract: Group communication, also known as multicasting, is a conceptually simple yet highly interesting phenomenon in computer networks. The communication pattern of multicasting is inherent to a whole range of applications, a brief overview of which, along with their requirements, I shall present. Furthermore, I shall outline the principles of native multicast and the protocols that implement it. I shall highlight the advantages, as well as the problems associated with this solution. I shall then examine virtual multicast as an overlay network implemented using active network elements. I will demonstrate the characteristics that an active element for implementing virtual multicast must fulfil, and under what circumstances it represents a better solution than the use of native multicast. I will analyse in more detail the specific development and implementation of the active element, including the results of performance measurements which demonstrate its usability even where native multicast fails. At the end of the lecture, I will then demonstrate the support that virtual multicast provides, using an active element, for the applications presented at the beginning of the lecture.

7 October 2008

Doc. RNDr. Luděk Matyska, CSc., FIMU, Brno

Large-scale distributed systems in the real world

Abstract: The field of parallel and distributed systems has a long tradition stretching back several decades. Truly large-scale distributed systems, which began to emerge in the last decade and whose development was made possible by the improving quality of internet connections, have raised a number of new problems, for which solutions are still being sought. In this lecture, I shall focus on three areas

• Monitoring the state of the system and, in particular, the tasks/processes running within it
• Task scheduling in a large-scale distributed environment
• Authentication and authorisation

where I will highlight both the problems and, in particular, their current solutions, as well as our contribution to this field.

Perhaps somewhat surprisingly, a number of problems, particularly in the areas of scheduling and security, can also be addressed in distributed systems through the virtualisation of computers. I will demonstrate which seemingly intractable problems can be circumvented in this way, and what impact consistent virtualisation – including virtualisation that interconnects networks – has on the development and usability of distributed systems. The findings presented in this lecture are largely based on experience gained from building large-scale distributed systems at both national and international levels. In conclusion, I shall therefore briefly touch upon the current state of large-scale distributed computing and storage infrastructures in the Czech Republic and the EU, including our specific contribution to this process. Large-scale distributed systems represent the next stage of development, the first step of which was the Internet. Understanding the problems associated with their practical design and operation, and contributing to their resolution, is one of the greatest challenges facing computer science in the coming years.

14 October 2008

RNDr. Tomáš Pitner, Ph.D., FIMU, Brno

Web 2.0 — Core Concepts, Applications, and Implications

Abstract: Web 2.0 represents the recent evolutionary shift towards a more user-oriented and user-driven web. The presentation will attempt to distil the essence of this concept — identifying the distinguishing characteristics of the new web, illustrating them through successful services, and commenting on the trends. As a specific topic, the phenomenon of integrated services – ‘mashups’ – will be discussed, highlighting its key issues, including technological, legal and business aspects.

21 October 2008

RNDr. Petr Sojka, Ph.D., FIMU, Brno

From Minds to Pixels and Back

Abstract: We will present several approaches and technologies in the areas of electronic document preparation. The first part will cover those that facilitate the flow of information from authors’ minds to the pixels on paper or screen (segmentation, hyphenation, etc.). The second part will reflect on work carried out in the field of digitising technical documents to help scholars make the most of scanned mathematical papers by returning them to their minds via a structured digital library.

4 November 2008

RNDr. Zdenko Staníček, Ph.D., Mgr. Pavel Minařík, FIMU, Brno

MyNetScope – a Service System for Computer Network Analysis and Monitoring

Abstract: Introduction to the fundamentals of the emerging field of Service Science, both in academic and business contexts. The concepts of Service and Service System. Service System architecture. From Universal Information Robots to solutions based on Mycroft technologies . A four-level framework for solution usability. Architecture of scope-like applications. MyNetScope—a solution for monitoring computer networks. MyNetScope and network security. MyNetScope and SSH password cracking attacks. Demonstration of current results from the MU network.

11 November 2008

Mgr. et Mgr. Jan Krhovják, RNDr. Jan Bouda, Ph.D., doc. RNDr. Václav Matyáš, M.Sc., Ph.D., RNDr. Petr Švenda, FIMU, Brno

Towards true random number generation in mobile environments

Abstract:

18 November 2008

Prof. Thilo Kielmann, Vrije University Amsterdam, The Netherlands

Coarse-grained Quantitative Modelling of C. elegans Vulval Development using Petri Nets

Abstract: Understanding the processes involved in multicellular pattern formation is a central problem in developmental biology, which will hopefully lead to many new insights, for example, in the treatment of various diseases. However, defining suitable computational models for development modelling, capable of performing in silico simulation experiments, remains an open and challenging problem. We propose a novel, coarse-grained, quantitative approach based on Petri nets. Our model retains the simplicity of the basic Petri net formalism, whilst focusing on the underlying biological processes and attempting to mimic their behaviour. As a case study, we apply our modelling approach to the well-studied C. elegans vulval development process. We show that our model correctly reproduces a large set of in vivo experiments, with statistical accuracy. It also generates gene expression time series consistent with recent biological evidence.

25 November 2008

Prof. RNDr. Petr Kůrka, CSc., CTS UK, Prague

Numbers and their names

Abstract: The lecture will discuss symbolic representations of real numbers (finite or infinite strings of letters), arithmetic algorithms in these representations and computer arithmetic.

2 December 2008

Prof. Ing. Václav Hlaváč, CSc., Faculty of Electrical Engineering, Czech Technical University, Prague

Pose-primitive-based human action recognition in videos or still images (joint work with Christian Thurau)

Abstract: This talk will present a method for recognising human actions based on pose primitives. In the learning mode, the parameters representing poses and activities are estimated from videos. In the run mode, the method can be used for both videos and still images. To recognise pose primitives, we extend a Histogram of Oriented Gradients (HOG)-based descriptor to better handle articulated poses and cluttered backgrounds. Action classes are represented by histograms of pose primitives. For sequences, we incorporate the local temporal context using n-gram expressions. Action recognition is based on a simple histogram comparison. Unlike mainstream video surveillance approaches, the proposed method does not rely on background subtraction or dynamic features and thus enables action recognition in still images.

9 December 2008

Pavel Rychlý, MSc, PhD, FIMU, Brno

Even larger web corpora

Abstract: Text corpora are an important source of information for a range of areas in computational linguistics. A consequence of Zipf’s law is that the vast majority of phenomena in natural language occur sparingly, and we often do not have sufficient data available to study these rare phenomena. In this respect, more data means better data, and larger corpora are better corpora. Although an enormous amount of text is available on the web for some languages, the web corpora created to date have by no means exceeded the three-billion-word mark. This talk will outline our approach to creating a web corpus of English texts with a target size of twenty billion words. In describing the related problems and the tools used, we shall focus in particular on the detection of similar (duplicate) documents in collections of this scale and demonstrate our own effective solution to this problem.

16 December 2008

RNDr. Petr Matula, Ph.D., FIMU, Brno

Automated single-cell-based image analysis of high-throughput cell array screens for quantification of viral infection

Abstract: Systems biology is a relatively new interdisciplinary field of research that focuses on the systematic study of complex interactions in biological systems. High-throughput techniques are used to collect quantitative data for the construction and validation of models of the systems’ behaviour. A central issue in image-based high-throughput screening is the efficient, robust and automated single-cell-based analysis of massive image datasets.

This talk concerns an siRNA-based screening system using cell arrays and high-throughput fluorescence microscopy for the quantification of viral infection. We have developed this system at the University of Heidelberg and the German Cancer Research Centre, Heidelberg, Germany.