Our seminar

Laboratory seminars (PV273 in the course catalog) on Wednesday, 15:00-16:00, A505, FI MU, Botanická 68

The format of standard lectures: 30 minutes presentation + 15 minutes for questions, slides in English, presentation in English or Czech based on audience

Programme for Spring 2019

  • 20.2.2019
    RNDr. Petr Ročkai, Ph.D.
    Software verification vs security vulnerabilities
    Abstract: Applications of formal methods (especially automated verification methods) to realistic software is an emerging field with an active research community. Programs written in C are especially interesting since the C language is still widely used in system-level programs, while also being susceptible to many serious programming mistakes. Additionally, in many cases, such mistakes are hard to uncover using traditional methods like testing. Formal methods attack the problem from a different angle, using powerful automated reasoning to discover hard-to-find bugs. It is perhaps not surprising that a substantial overlap exists between security vulnerabilities and the classes of bugs that can be exposed using such tools. In this talk, we will explore the state of the art in software verification from a security perspective.
  • 27.2.2019
    doc. RNDr. Aleš Horák, Ph.D.
    Anonymous authorship and propaganda detection in texts
    Abstract: The joint project of researchers from FSS, FI and PrF MU named „Manipulative Techniques of Propaganda at the Time of the Internet“ responds to the current issue of spreading political propaganda of foreign actors through new media and social networks in the Czech Republic.
    One of the outputs of the project will be an analytical software for the detection of manipulative techniques in the text. The employed methods lean on previous experience with developing a system for recognizing anonymous authorship. This discipline focuses on the tasks of authenticating documents in court cases and searching for authors of anonymous illegal documents in the Internet environment. With support from the Ministry of the Interior, we have developed the Authorship Recognition Tool (ART) to determine the authorship of documents using the author stylistic analysis. ART solves the problems of verifying and assigning authorship and document clustering by the (presumed) author.
  • 6.3.2019
    Mgr. David Střelák
    Performance analysis and autotuning setup of the cuFFT library
    Abstract: Fast Fourier transform (FFT) has many applications. It is often one of the most computationally demanding kernels, so a lot of attention has been invested in tuning its performance on various hardware devices. However, FFT libraries have usually many possible settings, and it is not always easy to deduce which settings should be used for optimal performance. In practice, we can often slightly modify the FFT settings, for example, we can pad or crop input data. Surprisingly, a majority of state-of-the-art papers focus to answer the question of how to implement FFT under given settings but do not pay much attention to the question which settings result in the fastest computation.
    We target a popular implementation of FFT for GPU accelerators, the cuFFT library. We analyze the behavior and the performance of the cuFFT library concerning input sizes and plan settings. We also present a new tool, cuFFTAdvisor, which proposes and using autotuning finds the best configuration of the library for given constraints of input size and plan settings. We experimentally show that our tool can propose different settings of the transformation, resulting in an average 6× speedup using fast heuristics and 6.9× speedup using autotuning.
  • 13.3.2019
    RNDr. Michal Procházka, Ph.D.
    ELIXIR AAI: Building large pan-European authentication and authorization infrastructure
    Abstract: Several years ago, ELIXIR CZ (represented by Institute of Computer Science and its Centre CERIT-SC) together with ELIXIR FI (represented by CSC) proposed to build and operate an authentication and authorization infrastructure (AAI) for the large pan-European research infrastructure ELIXIR. The ELIXIR AAI is already running in production for two years and is one of the most advanced AAI in the academic area. Its evolution propagated back to the designs and schemas for general AAI, and it even aims to become inspirational for the EOSC AAI to glue together research infrastructures. The experience gained during the development and operation of ELIXIR AAI will be used for the evolution towards the Life Science AAI that will serve for all the European LS infrastructures. The concepts and achievements are also published in an extensive F1000 paper and materialized in the AARC Blueprint architecture. The presentation will describe the way how we started with our identity and access management system Perun and got to a complete AAI.
  • 20.3.2019
    Mgr. Tomáš Sapák
    Delivering Single Sign-on at MUNI with Docker and Puppet
    Abstract: Single Sign-on at MUNI is a service, that provides user with secure authentication to wide variety Masaryk University and international web services. Service itself consists of a few interconnected software components (web server, application server, database, directory service, caching service) accessed by several thousand unique users each day. Maintaining complex system with security and high availability requirements brings a few challenges – accountability, system state definition, frequent automatized patching, OS upgrades, handling software dependencies and documentation accuracy. This presentation will show the audience, how we solved these issues by moving from traditionally installed applications to system running on top of docker containers continuously delivered by Puppet.
  • 27.3.2019
    Future Sitola
  • 3.4.2019
    Mgr. Filip Münz, PhD.
    Planning and evaluation of experiments (40 minuts for presentation)
    Abstract: The standard analysis of variance is usually applied to categorical data; however, the method is principally an analog to subtraction of a parametric model that also attempts to „explain out“ the most of observed variance. In the case of highly-dimensional data, the process can be extended by the method of principal components that unfortunately often blurs our understanding of the model for the sake of reducing the amount of unexplained variability. Another tradeoff to consider is the bias vs. variance decomposition linked to the number of model parameters. The importance of model validation can be demonstrated in this case.
    Slides with source codes
    Video from the presentation
  • 10.4.2019
    RNDr. Michal Zima
    Mass mining of cryptocurrencies
    Abstract: Cryptocurrencies were popularised a lot through their process of “mining.” There are various approaches to mining with different pros and cons. Different cryptocurrencies choose different ones. Some even cooperate and share their miners. This approach (called “merged mining”) seems to be increasingly popular in use but is not exploiting its full potential yet.
    In my talk, I will initiate you into the mysteries of cryptocurrency mining and present the state of the art of merged mining, both its advantages and issues, my recent work in this area and opportunities that this approach provides, mainly in enhancing security of involved cryptocurrencies.
  • 17.4.2019
    Mgr. Luděk Matyska
    Quantum computers of today
    Abstract: The technological advancement and implementations in recent years allowed researchers and companies to commercially offer devices supposedly capable of quantum computation as either cloud services or physical computational units. We assess these commercially available devices from several companies (IBM, D-wave Systems, etc.) by looking at their physical implementations, their performance and capacity, universality, and how much can the customer trust something exploiting the laws of quantum mechanics is actually happening inside.
    The second part of the talk is then a closer look at the Q Experience, a cloud service offering access to the universal quantum computer, as a quick introduction into a very intuitive graphical way of programming a quantum computer.
  • 24.4.2019
    UVT seminar (no presentation)
  • 15.5.2018
    CESNET seminar (no presentation)
  • 22.5.2019
    Mgr. et Mgr. Jaroslav Oľha
    Autotuning in complex systems and its applications
    Abstract: Many problems arise when the computational load of a single scientific simulation is distributed among multiple processors and accelerators – chief among them being hardware heterogeneity and performance portability.
    There are two successful approaches that tackle these problems dynamically: task-based runtime systems and autotuning. While task-based runtime systems treat various parts of the application code as separate tasks and schedule them across the distributed hardware resources, autotuning tries to optimize parts of the code for better performance.
    Since these approaches are clearly complementary, we hypothesise that their combination could result in significant performance gain in heterogeneous distributed systems. We intend to test our hypothesis by implementing an autotuning task-based system, and applying it in a real-life image-processing pipeline in cryogenic electron microscopy.

Past seminars

Contact: Hana Rudová

(c) 2018 SITOLA, Administration