![\"Our](\"https:\/\/www.sitola.cz\/wordpress\/wp-content\/uploads\/2018\/04\/seminar-180418-long-s-3.png\")
<\/p>\n<\/p>\n
This is the programme of the Sitola seminar in autumn 2020. Presentations for the current semester are available here<\/a>.<\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n <\/p>\n Pavel Dost\u00e1l: Planning for high-speed railways<\/a> <\/p>\n <\/p>\n<\/ul>\n","protected":false},"excerpt":{"rendered":" This is the programme of the Sitola seminar in autumn 2020. Presentations for the current semester are available here. 7.10.2020 RNDr. Ji\u0159\u00ed Filipovi\u010d, Ph.D. Using hardware performance counters to speed-up autotuning convergence on GPUs Abstract: Nowadays, GPU accelerators are commonly used to speed up general-purpose computing tasks on a variety of hardware. However, due to […]<\/p>\n","protected":false},"author":42,"featured_media":0,"parent":2896,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"_links":{"self":[{"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/pages\/3550"}],"collection":[{"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/users\/42"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/comments?post=3550"}],"version-history":[{"count":4,"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/pages\/3550\/revisions"}],"predecessor-version":[{"id":3554,"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/pages\/3550\/revisions\/3554"}],"up":[{"embeddable":true,"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/pages\/2896"}],"wp:attachment":[{"href":"https:\/\/www.sitola.cz\/wordpress\/wp-json\/wp\/v2\/media?parent=3550"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}\n
\nRNDr. Ji\u0159\u00ed Filipovi\u010d, Ph.D.
\nUsing hardware performance counters to speed-up autotuning convergence on GPUs<\/strong>
\nAbstract: Nowadays, GPU accelerators are commonly used to speed up general-purpose computing tasks on a variety of hardware. However, due to the diversity of GPU architectures and processed data, optimization of codes for a particular type of hardware and specific characteristics of processed data can be extremely challenging. The autotuning of performance-relevant source-code parameters allows to tune applications automatically and keeps the performance portable. Although the autotuning process typically results in speeding-up the code, it can bring unacceptable overhead if (i) tuning space is vast and full of poorly-performing implementations, or (ii) the autotuning process has to be repeated frequently because of changes in processed data or migration to different hardware.
\nWe introduce a novel method for searching tuning spaces. The method takes advantage of collecting hardware performance counters (also known as profiling counters) during empirical tuning. Those counters are used to bias the searching process towards faster implementations. The method requires a sample of tuning space to be observed at any GPU. It builds a problem-specific model, which can be used during autotuning of the full tuning space on various, even previously unseen inputs or GPUs. We experimentally demonstrate that our method can speed-up autotuning when the application has to be ported to the different hardware or when it has to process data of different characteristics.\n<\/li>\n
\nMgr. Martin \u010cech
\nThe most interesting bits of a giant scientific open-source project<\/strong>
\nAbstract: The scientists in biology, physics, astronomy, climate change, and other fields are relentlessly producing exabytes of data. Analysis of such is a major endeavor oftentimes not only from the novelty standpoint but also from the software and infrastructure management angle. Galaxy Project is a web-based platform for computational research that helps hundreds of thousands of scientists address software pipeline reproducibility and accessibility challenges. More than 150 public and thousands of private instances worldwide leverage heterogeneous infrastructures from personal laptops, through cloud computing, to national HPC centers.
\nIn this talk, I will present the most intriguing parts about this undertaking, which includes an open global delivery network of petabytes of research data using the CernVM File System, challenges of working on a flat-hierarchy distributed team including most of the world\u2019s time zones, Europe-spanning shared compute node network that anybody can attach to and other efforts of what Open Hub names \u201cone of the largest open-source teams in the world\u201d.\n<\/li>\n
\nMgr. Adam Janovsk\u00fd
\nOn large-scale analysis of cryptographic API in Android malware<\/strong>
\nAbstract: Cryptography has been extensively used in Android applications to guarantee secure communications, to conceal critical data from reverse engineering, or to ensure the privacy of mobile users. Various system-based and third-party libraries for Android provide cryptographic functionalities, and previous works especially explored the misuse of such functionalities in benign applications. However, cryptographic API has not yet been explored in Android malware. In this talk, we propose the first large-scale analysis of cryptographic API in Android malware. In particular, we analyzed 251 996 malicious applications released between 2012 and 2018, gathering nearly 1 million of cryptographic API expressions. Our results reveal what cryptographic primitives are employed by the malware authors, providing a valuable asset for the area of threat intelligence. For instance, we discovered widespread use of weak hash functions, the growth of public-key cryptography, and the progressive decrease of the use of cryptographic API in malware. The attained results can help the researchers and the security community to shed light on the skills and aims of malware creators. The directions of the future work will be outlined that leverage the large volume of acquired data to learn more about the misuse of cryptographic API by malware creators.\n<\/li>\n
\nRNDr. David St\u0159el\u00e1k
\nExperiences from complex software development<\/strong>
\nAbstract: It all starts with a single main(). Then, a bit of extra functionality is added. And before you know it, you are looking at a multilanguage project consisting of hundreds of thousands of lines of code from multiple contributors.
\nIn I2PC, we are working on two big projects for cryo-electron microscopy: Xmipp, which is a collection of (standalone) programs, each performing a specific task of the processing pipeline, and Scipion, which is a framework encapsulating multiple software plugins into a unified user experience. Currently, we have over 33 people managing over 350k+ lines of (core) code in 25 repositories.
\nIn this presentation, we will look at best practices and tips that you should keep in mind while working on such a big project, as long as you aim to deliver a fast and correct solution. More specifically, we will look at testing (why and how) and means of performance optimization, with additional focus on Python and C++.<\/p>\n
\nMachine learning at Sitola<\/strong>
\nAle\u0161 K\u0159enek: Simplistic but efficient use of neural networks in scientific computations: how (and why) it works?
\nPetr Holub, Tom\u00e1\u0161 Br\u00e1zdil: Machine learning on digital pathology and health data
\nJi\u0159\u00ed Filipovi\u010d: Machine learning in autotuning
\nTom\u00e1\u0161 Rebok: Machine learning for cell deaths detection<\/p>\n
\nRNDr. Tom\u00e1\u0161 Rebok, Ph.D.
\nANALYZA Platform for large-scale heterogeneous data analysis<\/strong>
\nAbstract: Large-scale heterogeneous data processing and mining are significant challenges in the Big Data analysis domain. High variability in data types and formats and analyses performed through different data domains pose significant challenges, especially when dealing with big amounts of data. Thus, the application of modern data processing techniques is required, which have to be further supported by powerful infrastructure allowing for data integration and efficient and flexible analysis. During our talk, we present the ANALYZA platform (developed by the ANALYZA project), which addresses these issues in the Police investigation domain. Besides its crucial architecture concepts, we also present a few use-cases demonstrating its application for a unified analysis of large-scale complex data (even in different data domains).<\/p>\n
\nSoumen Atta, Ph.D.
\nRecent research in dynamic vehicle routing<\/strong>
\nAbstract: Over the years, planning and scheduling operations of transport systems in modern logistics have received much attention. As a result, the classical vehicle routing problem has seen significant changes. Dynamic changes are frequently studied to consider modern-day logistics requirements. In this talk, a brief review of the works done in the last decade in the field of the dynamic vehicle routing problem (DVRP) is discussed. After the introduction of vehicle routing problems, the talk will concentrate on their dynamic characteristics. The major solution methodologies for solving DVRP will be discussed together with the complexity and scale of the problems solved nowadays.<\/p>\n
\nBc. Martin Piatka
\nOpen-source implementation of low-latency, high-definitiion 360 video transmissions<\/strong>
\nAbstract: 360-degree video is more immersive than classical video. However, capturing and playing is also more complex and computationally demanding. This presentation talks about developing a CUDA accelerated 360-degree video stitching library used in UltraGrid to enable low-latency high-definition streaming of 360 video. The talk also mentions playback of 360 video using a virtual reality headset.<\/p>\n
\nSearch at Sitola<\/strong><\/p>\n
\nDiscussion about future Sitola and
\nChristmas meeting<\/strong>\n<\/li>\n
\nRNDr. Matej Pivoluska, Ph.D.
\nRecent research in high-dimensional Quantum key distribution<\/strong>
\nAbstract: The discovery of an efficient algorithm to factorize large integers on quantum computers undermined the foundations of public key cryptography. With the current rapid development of quantum computers, the need to deploy cryptographic protocols immune to quantum attacks is more pressing than ever. Luckily, quantum mechanics also offers a solution to this problem in the form of Quantum key distribution protocols. It allows two users to exchange random and secret keys, which can be subsequently used in secret key cryptographic protocols. Since QKD’s security is based on physical laws (instead of computation complexity arguments), it is naturally immune to quantum attacks.
\nAlthough Quantum key distribution is arguably the most mature application of quantum information science, there are still two big challenges left: relatively small noise robustness (i.e., possibility of exchanging keys even in noisy environments) and relatively low key rates. The use of infinite dimensional degrees of freedom to address these problems has been proposed already at Millenium’s beginning. However, a significant increase in noise-robustness was never achieved in practice. In our recent paper, we propose a new QKD protocol that achieves high-noise robustness and key rates. In another recent paper, we also demonstrate the practical viability of our protocol and build a proof of principle implementation based on path degree of freedom.<\/p>\n
\nRNDr. Luk\u00e1\u0161 Ru\u010dka
\nPB176 – Basics of quality and management of source code<\/strong>
\nThe faculty has recently introduced a new study program (Programming and application development), intending to produce junior and senior developer ready graduates. However, a weak point in graduate competencies, often reported both inside and outside of the faculty environment – the knowledge of essential code management and quality assurance tools and workflows – is still an issue. The newly created bachelor course shall address the issue in five topic blocks, three invited talks of QA staff from distinct company sizes, and one Kobayashi Maru overnight exercise. Although the course will launch this spring, it is so far planned without seminars.
\nTherefore, I would like to invite the audience to familiarize themselves with the course and comment on (either parallel or follow up) seminar options and perhaps the course as a whole.<\/p>\n
\nMeetings devoted to state exam rehearsals are scheduled mostly. State exam rehearsals are in Czech mostly. <\/p>\n
\nState exam rehearsals (master)<\/strong>
\nMartin Piatka: N\u00edzkolaten\u010dn\u00ed p\u0159enosy 360 stup\u0148ov\u00e9ho videa<\/a>
\nVedouc\u00ed: Milo\u0161 Li\u0161ka
\nOponent: Martin Pulec\n<\/li>\n
\nState exam rehearsals (master) <\/strong>
\nVojt\u011bch Sassmann: Vehicle routing with metaheuristics<\/a>
\nVedouc\u00ed: Hana Rudov\u00e1
\nOponent: Tom\u00e1\u0161 Br\u00e1zdil<\/p>\n
\nVedouc\u00ed: Hana Rudov\u00e1
\nOponent: Barbora B\u00fchnov\u00e1\n<\/li>\n
\nState exam rehearsals (bachelor) <\/strong>: cancelled due to illness
\nMikol\u00e1\u0161 Stuchl\u00edk: Idiomatick\u00e9 rozhran\u00ed jazyka Swift pro pr\u00e1ci s knihovnou GTK+<\/a>
\nVedouc\u00ed: Michal Zima
\nOponent: Luk\u00e1\u0161 Ru\u010dka\n<\/li>\n