Once a week during the academic year, an invited speaker (from abroad, as well as from the Czech Republic) talks about his or her scientific work. Colloquium takes place at the Faculty of Informatics and is open to the scientific community. Lecture dates can be found in the programme. Tuesday 14.00 - 15.00, D2, FI MU, Botanická 68a

Colloquia programme with abstracts for the Spring 2019 semester

26. 2. 2019
RNDr. Kateřina Schindlerová, CSc., Faculty of Computer Science, University of Vienna
Causality in Time Series: A Short Introduction and Current Challenges
Abstract: We explain the concept of quantitative causality by various approaches and focus on Granger causality. Causality among a higher number of processes by means of graphical Granger causality will be presented and its applications shown. Current questions and open problems will be sketched from the point of view of heterogeneous and big data.
5. 3. 2019
Mouzhi Ge, Ph.D., FI MU
Research Paradigms in Computer Science, Information Systems, and Design Science
Abstract: There are different research paradigms in informatics, three of the typical research paradigms are Computer Science, Information Systems and Design Science. In this talk, I will discuss the three research paradigms, including their methodological features, research communities, and how to conduct research in and across different research communities. Then I will demonstrate the research paradigms by research publications and interactive discussions. The target audience of this talk is more expected to be Ph.D. students and junior researchers.
12. 3. 2019
Ing. Tomáš Mikolov, Ph.D., Facebook AI Research
Recent progress in AI research
Abstract: The last decade did bring us many new and improved applications of machine learning. To name a few, we have seen dramatic improvements in web search, user post classification, statistical machine translation, speech recognition, and image classification. But how far are we from the grand goal, which is the development of truly general AI?

In this talk, I will present some of the ideas that I find promising for the development of artificial intelligence. These include a discussion about artificial life, dynamical systems, complex adaptive systems, and mathematical models of evolution.

19. 3. 2019
RNDr. Radek Ošlejšek, Ph.D., FI MU
Semantic data modeling – conceptual view
Abstract: In this talk, I will explain the principles of semantic modeling. I will briefly compare popular data models, such as rational vs. NoSQL databases. Then, I will focus on graph ontologies. I will describe their features, the problem of abstraction, information levels (e.g., upper vs. domain ontologies), and other aspects that are important for semantic modeling. The discussion will be supported by practical examples that emerged in our research of the semantic modeling of image content.

Examples from the cybersecurity domain will, in turn, demonstrate the abilities and limitations of ontologies for detection and mitigation of cyber threats. Possible utilization of ontologies for formal knowledge modeling in hands-on cybersecurity training will be presented as well.

26. 3. 2019
RNDr. Petr Švenda, Ph.D., FI MU
Analysis and use of RSA keypair generation bias
Abstract: The talk will explain the details of recently discovered bias in RSA keys produced by various cryptographic libraries. The bias allows to attribute RSA public key to its origin library and already lead to results with practical security impact: 1) A discovery of factorable keys from smartcards (so-called ROCA vulnerability), 2) a revelation of cryptographic keys supposedly generated on-chip while actually generated outside and only later injected and 3) a first direct measurement of cryptographic smartcard popularity in Internet-wide datasets. The reason for the bias presence steaming from the prime generation algorithms will be explained with typical usage scenarios for accurate key origin attribution. The talk will also cover ongoing research and open research directions.
2. 4. 2019
prof. Mgr. Michal Koucký, Ph.D., MFF UK
Approximating edit distance within constant factor in sub-quadratic time
Abstract: Edit distance is a measure of similarity of two strings based on the minimum number of character insertions, deletions, and substitutions required to transform one string into the other. It has numerous applications in various fields. The edit distance can be computed exactly using a dynamic programming algorithm that runs in quadratic time. Andoni, Krauthgamer and Onak (2010) gave a nearly linear time algorithm that approximates edit distance within approximation factor poly(\log n). In this talk I will present an algorithm with running time O(n^{2-2/7}) that approximates the edit distance within a constant factor.

Based on a joint work with Diptarka Chakraborty, Debarati Das, Elazar Goldenberg, and Mike Saks.

9. 4. 2019
Prof. Dr. rer. nat. habil. Timo Ropinski, Faculty of Engineering, Computer Science and Psychology, Ulm University
Interactive Visualization Techniques for Molecular Structures
Abstrakt: Within this talk I will cover our work in the area of molecular visualization, and will present techniques which have been developed with the goal to improve the spatial comprehension of complex molecular structures. First, coverage-based opacity estimation is discussed as a technique to achieve Depth of Field (DoF) effects when visualizing molecular structures. The proposed algorithm is an object-based approach which eliminates many of the shortcomings of state-of-the-art image-based DoF algorithms. Based on observations derived from a physically-correct reference renderer, coverage-based opacity estimation exploits semi-transparency to simulate the blur inherent to DoF effects. Second, I will discuss the integration of diffuse illumination effects into molecular visualization. While current molecular visualization techniques utilize ambient occlusion as a global illumination approximation in order to improve spatial comprehension, interreflections are also known to improve the spatial comprehension of complex geometric structures. To realize these interreflections in real-time, an analytic approach is exploited for capturing interreflections of molecular structures. By exploiting the knowledge of the underlying space filling representations, the required parameters can be reduced and symbolic regression can be applied to obtain an analytical expression for interreflections. I will discuss how to obtain the data required for the symbolic regression analysis, and how to exploit the analytic solution to enhance interactive molecular visualizations. Finally, I will show how this approach can be extended to other molecular representations.
16. 4. 2019
Dr. rer. nat. RNDr. Mgr. Bc. Jan Křetínský, Ph.D., Department of Informatics, Technical University of Munich, FI MU
Machine learning in verification
Abstract: On the one hand, formal verification methods provide hard guarantees on analysis results, but do not scale well and are often hard to use. On the other hand, machine learning comes with weak or no guarantees, but scales well and can provide more understandable solutions. In this talk, we show several examples how these approaches can be combined and the best of the two worlds achieved. We demonstrate this on controller synthesis and controller representation in the setting of Markov decision processes and games.
23. 4. 2019
RNDr. Vojtěch Řehák, Ph.D., FI MU
What is the best timeout?
Abstract: We all have some experience with deadlines, and we know how hard it is to deliver results in time, especially when some obstacles with an uncertain solving time may occur. In this talk, we deal with formalisms suitable for modeling and analysis of such systems, i.e., reflecting both timeouts and actions taking stochastic continuous time. We show how this kind of stochastic timed systems can be modeled and what properties can be computed for such models. On surprisingly small examples we also demonstrate some fundamental limits of such models. Then we will consider that some parts of the model are unknown and we take them as parameters that we would like to synthesize automatically. Finally, we will present results allowing for the synthesis of (epsilon)-optimal timeouts in such systems.
30. 4. 2019
Mgr. Alžběta Krausová, LL.M., Ústav státu a práva AV ČR
Legal and Ethical Aspects of Developing AI and Its Applications
Abstract: AI is a groundbreaking technology that raises a number of serious questions from both ethical and legal point of view. The lecture will introduce the main ethical concerns related to AI development as well as the EU guidelines on how to design intelligent applications in compliance with ethical requirements. Moreover, since ethical approach is conditioned by compliance with laws, the lecture shall also introduce the main legal requirements on AI systems and potential related legal problems including liability for AI, data processing and intellectual property protection. The lecture aims to provide an overall overview of complex legal problems with individual examples in order to highlight what needs to be taken in account while designing, developing, and running smart applications.
7. 5. 2019
prof. RNDr. Roman Barták, Ph.D., MFF UK
Multi-Agent Path Finding on Real Robots
Abstrakt: Multi-agent path finding (MAPF) deals with the problem of finding a collision-free path for a set of agents (robots). An abstract model with a graph describing the environment and agents moving between the nodes of the graph has been proposed. This model is widely accepted by the MAPF community, and the majority of MAPF algorithms rely on this model. In this talk, we argue that the model may not be appropriate when the plans are to be executed on real robots. We provide some empirical evidence that abstract plans deviate from real plans executed on robots and we compare several variants of abstract models. The talk motivates further research on the abstraction of problems concerning the applicability of solutions in practice.
14. 5. 2019
doc. RNDr. Aleš Horák, Ph.D., FI MU
prof. PhDr. Karel Pala, CSc., FI MU
Karel Pepper, the robot
Abstrakt: In the talk, we will introduce the new social robot by Softbank Robotics denoted as Pepper. We will present the robot hardware capabilities as well as first examples of natural human-machine dialogs in Czech, which are being developed by the team at FI MU. The robot is designed for social interactions with people and it is equipped with an extensive API set to detect faces, mood, or age and to react to their values. We will also sketch our plans to use the robot in both students works and in research with applications oriented to human-machine cooperation.
21. 5. 2019
prof. RNDr. Jiří Zlatuška, CSc., FI MU
Seeing the invisible - black holes
Abstract: Black holes used to be just an astrophysicsl curiosity representing extremal conditions following from General Ralativity. Until recently, no direct observation would seem feasible. Advanced use of computing allowed spectacular changes in this, resulting especially in the first image of black hole in the center of M87 using a technique resulting in computer-synthesized telescope stretching accross the diameter of the Earth as well as synthesizing a real image of black hole, not just a result of its simulation. In the talk a quite detailed and popular explanation will be provided how this remarkable scientific outcome has been obtained.
24. 9. 2019
RNDr. David Šafránek, Ph.D., FI MU
Computer-Aided Systems Biology
Abstract: In this lecture, we will introduce a novel paradigm for solving problems in biology. First, we will describe the framework of systems biology. Second, we will present an original approach to systems biology implemented on fundamental principles developed in computer science. Part of the talk will be dedicated to the comprehensive modelling platform we are developing in our research.