Translated using 
Machine-translated page for increased accessibility for English questioners.
Once a week during the academic year, a guest lecturer (from abroad or from the Czech Republic) gives a talk on their research. The colloquium, which takes place at the Faculty of Informatics, is open to the academic community. You can find the lecture dates in the programme.
Tuesday 14.00–15.00, D2, FI MU, Botanická 68a
Colloquium programme with abstracts for the Spring 2013 semester
- 19 February 2013
- Prof. Melius Weideman, Faculty of Informatics and Design, Cape Peninsula University of Technology, Cape Town, South Africa
- A successful website = Visibility + Usability: www.muni.cz?
- Abstract:
The objective of this colloquium is to identify, analyse and list the underlying factors determining the overall success of a website. This approach contrasts with some traditional computer science perspectives, such as technical platforms, programming, hardware or web services. The case of www.muni.cz is briefly examined.
As an introduction, the background and fundamentals of a typical website are outlined:
- - the key players,
- - search engine rankings,
- - crawlers and results,
- - the role of SEO and PPC,
- - user behaviour, and
- - defining and measuring website visibility and usability.
- 26 February 2013
- Prof. RNDr. Jiří Zlatuška, CSc., Faculty of Informatics, Masaryk University
- Libraries in the Age of Digitalisation and Open Access
- Abstract: The digitisation of scholarly publications is leading to a significant change in the role of academic libraries and the technical infrastructure associated with them (in its most simplistic form, for example, the question of the future of libraries in an era when paper publications become a thing of the past). Open Access and institutional repositories making scientific literature accessible represent not only a necessary change in the activities of contemporary scientific institutions and the dissemination of their research findings, but also new opportunities and challenges for applied research combining library science and information technology.
- 5 March 2013
- Dr Vashti Galpin, Ph.D., LFCS, School of Informatics, University of Edinburgh
- Formal modelling of biological systems
- Abstract: Process algebras are formalisms that describe the concurrent behaviour of systems. In recent years, they have been applied to the modelling of biological systems. Bio-PEPA is a quantitative process algebra used for this type of modelling and is based on the stochastic process algebra PEPA, which was developed for modelling the performance of human-created systems such as computer networks. In this presentation, the syntax and semantics of Bio-PEPA will be introduced, along with the various analysis techniques that are available for Bio-PEPA models. A model of protein trafficking within the mammalian cell and related results will be described, and, time permitting, semantic equivalences for Bio-PEPA will be briefly discussed. This is joint work with Jane Hillston.
- 12 March 2013
- Associate Professor Dalia Krikščiūnienė, FI MU, Kaunas Faculty of Humanities, Vilnius University
- Anomaly detection methods and applications
- Abstract:
Numerous statistical, econometric and intelligent methods are being researched in scientific work and industrial applications to identify recurring patterns and approximate distribution rules in data series. The main challenge faced by scientific research is the lack of regularity and the rather chaotic nature of real-world data. This hinders the performance of forecasting methods based on identifying regularities in data distribution, limits their application to the analysis of historical data repositories, and reveals the high unreliability of models in responding to external impacts in real-time data streams.
This research aims to explore the application of computational analysis methods for detecting behavioural anomalies. The methods and principles of anomaly detection are analysed from a theoretical perspective and evaluated for their ability to support analysis, forecasting and the design of optimal strategies for business operations across various application domains. These include the analysis of anomalies in financial markets caused by changes in investor activity due to various calendar effects, the impact of media announcements and news, and fluctuations in market conditions, including crises and bubbles. Anomaly detection is important in the fields of facility management based on sensor data systems, the detection of network intrusions and performance failures.
The results of recently published research cover several methods for detecting behavioural anomalies, including methods for evaluating information efficiency (Shannon’s entropy, Hurst exponent), the impact of events analysed using interrupted time series methods, and the evaluation of binary clustering algorithms.
The research results are assessed by discussing the potential of the methods explored for detecting behavioural anomalies within the expanding research field of Big Data.
- 19 March 2013
- Doc. RNDr. Petr Hliněný, Ph.D., Faculty of Informatics, Masaryk University
- Faster Algorithms for Well-Structured Graph Classes
- Abstract: The theory of parameterised complexity primarily studies computationally hard problems for which much faster (e.g. FPT) algorithms exist for special instances. Quite often, such simpler instances are characterised by their specific structure, expressed in terms of a suitable structural width parameter, and the faster algorithms utilise dynamic programming. We present a Myhill–Nerode-type formalisation of such algorithms and related algorithmic results.
- 26 March 2013
- RNDr. Tomáš Brázdil, Ph.D., Faculty of Informatics, Masaryk University
- Analysis of counter games
- Abstract: Counter-based games are a natural model of controllable systems that operate with various resources (e.g. fuel or battery power ) in an antagonistic environment. The aim of the formal analysis of such a system is to synthesise a control mechanism that ensures the resources are not depleted and that the system exhibits the expected functionality. A counter game represents resources using a set of non-negative integer counters. The control mechanism and the system’s environment correspond to two antagonistic players who, through repeated choices of moves, alter the values of the counters and the internal state of the system. Formally, counter games are a special case of two-player turn-based games on directed graphs with an infinite number of vertices. The lecture will describe the basic properties of these games, together with an analysis of the complexity of their solutions. Furthermore, special subclasses of games will be presented which preserve important properties from the perspective of system modelling whilst also allowing for efficient analysis.
- 2 April 2013
- Doc. RNDr. Lubomír Popelínský, Ph.D., Faculty of Informatics, Masaryk University
- Educational data mining
- Abstract: Educational data mining (EDM) aims to analyse any data generated during a learning or tutoring process, such as the evaluation of a student by a tutor, students’ behaviour whilst learning, or their progress through an e-learning course. We briefly outline the history of EDM and the most significant strands of EDM research, both past and present. We then focus on two projects carried out at the KD lab: predicting student drop-out based on their social behaviour (in collaboration with the IS MU team), and graph mining for the automatic classification of logical proofs. We conclude with a brief outlook to the future.
- 9 April 2013
- Dr Geraint Price, Royal Holloway, University of London
- Is Cyber Security any different from Information Security?
- Abstract:
In this presentation, we will discuss a number of key themes centred on
the notion of ‘Cyber Security’. We begin by asking whether Cyber
Security is simply more of the same, or whether there are genuine
differences compared to previous concepts such as ‘Information Security’ or
‘Information Assurance’.
Our early findings from the “Cyber Security Club” (an industrial research forum hosted at Royal Holloway) suggest that if we are to provide security that is fit for purpose in the 21st century, then we need to broaden our way of thinking and develop a more integrated response to the threats we face. There is also a need for new paradigms to tackle the challenges we face.
Finally, we highlight the need to cast our net more widely, forming stronger relationships with other disciplines – psychology, economics, sociology, amongst others – and learn from their experiences in dealing with problems analogous to those we face.
- 16 April 2013
- Doc. RNDr. Stanislav Tokár, CSc., Faculty of Mathematics, Physics and Informatics, Comenius University, Bratislava
- State of the art and goals of the ATLAS/LHC project at CERN
- Abstract: A retrospective survey of concepts in particle physics from the early 20th century to the present day and their implications for our understanding of the world around us. The LHC accelerator complex – what can be expected and why we need it.
- 23 April 2013
- Dr Adam Kilgarriff, Lexical Computing Ltd., Brighton, United Kingdom
- How to evaluate a corpus
- Abstract: Linguistics researchers and language technologists often ask themselves: ‘Which corpus should I use, or should I build one of my own? If I build one of my own, how will I know if I have done a good job?’ At present, there is very little guidance available to them. They require a framework for evaluating corpora. We have developed such a framework, specifically for corpora designed to provide good coverage of ‘general language’. The task we have set ourselves is the automatic creation of a publication-quality collocations dictionary. For a sample of 100 headwords in Czech and 100 in English, we identify a gold standard dataset of (ideally) all the collocations that should appear for these headwords in a collocations dictionary. We then use these to determine precision and recall for a range of corpora, with a range of parameters.
- 30 April 2013
- Prof. Štefan Porubský, DrSc., Institute of Computer Science, Czech Academy of Sciences
- A Tribute to Paul Erdős
- Abstract: To mark the 100th anniversary of P. Erdős’s birth, we shall commemorate his life, selected achievements, background and the roots of the Hungarian ‘mathematics machine’.
- 14 May 2013
- Dr Bruno Rossi, Centre for Applied Software Engineering, Free University of Bozen-Bolzano, Italy
- Emergent Design in Agile Software Development Practices
- Abstract:
One of the pillars of Agile software development practices is the
incremental nature of the development process, in which new
functionalities are developed on top of previous ones, iteration by
iteration. As such, architectural decisions form part of what is known as
‘emergent design’, whereby the overall system architecture emerges
over time.
In this talk, I will outline the emergent design process and contrast it with traditional up-front design. I will focus in particular on two key aspects: how do we ensure that the emergent design approach is suitable for our application or domain? How do we ensure that the emergent design process does not lead to mere chaos? In answering these questions, we must confront the widely held view of the traditional central role of the system’s architecture.
- 21 May 2013
- Doc. RNDr. Tomáš Skopal, Ph.D., Faculty of Mathematics and Physics, Charles University
- Axiom exploration for similarity indexing – beyond the metric space model
- Abstract: Database techniques for accelerating similarity search utilise a metric indexing model to partition the data space. For simplified representations of complex unstructured data, this model is sufficient; however, in modern applications, the requirement for the similarity function to be metric is becoming increasingly problematic. Current sporadic solutions for indexing non-metric similarities generally focus on mapping the problem onto a metric (or directly Euclidean) space. Whilst this approach to reducing the problem is elegant, it often fails to lead to an optimal solution. This manifests itself either as a high search error rate or a deterioration in indexability. In this talk, we will present a new paradigm which completely abandons the metric indexing model and aims to find models that are qualitatively different, yet at the same time tailored to the given similarities. This indexing model is sought by exploring the space of axioms suitable for indexing data, with this exploration taking place at the level of formulas, rather than, for example, through the numerical parameterisation of a single model. We shall show that the exploration of the space of axioms can be carried out both using rules and using genetic programming. The result of the entire process is then a set of axioms applicable to the indexing of databases under a specific similarity model. With a touch of hyperbole, we can view this method as a database analogue to the search for various physical laws applicable in different universes.