Seminar
Laboratory seminars (PV273 in the course catalog) on Wednesday, 10:30 – 11:30, A505, FI MU, Botanická 68, followed by an informal group lunch
The format of standard lectures: 30-40 minutes presentation + 15 minutes for questions, slides in English, presentation in English or Czech based on audience
- 5.9. – 7.9. 2024
SitSem seminar at Telč
See more about the program and some slides from Vizualization lectures - 25.9.2024
prof. PhDr. David Šmahel, Ph.D., RNDr. Ondřej Sotolář
Computational social science: Bridging social and computer sciences through machine learning
Abstract: We will present the work of the Interdisciplinary Research Team on Internet and Society (FSS & FI MU, irtis.muni.cz), which carries out research at the intersection of social sciences and informatics. First, we will present a brief summary of the different research areas that we are engaged in with our team. Next, we will explore the growing synergy between Natural Language Processing (NLP) and social science within the field of computational social science. We will examine how modern NLP techniques, based on representation-learning, can provide computational evidence for social science theories, enabling the analysis of large-scale textual data to study social behavior and communication. In turn, social science theories can inform and improve NLP models, offering insights into language use, context, and social structures. This feedback loop between NLP and social science fosters advances in both fields, enhancing the interpretation of data and strengthening theoretical frameworks. The lecture will show three recent papers from the IRTIS group as examples of this synergy. - 2.10.2024
Mgr. Václav Sobotka
Uncertainty in real-world vehicle routing
Abstract: Decades of research on vehicle routing problems have given rise to efficient and well-established heuristic methods. These methods can address rich-constrained industrial problems at scale while providing high-quality results reasonably fast. Unfortunately, the successful heuristics share a common silent assumption: all optimization inputs are assumed to be known precisely. Naturally, this assumption rarely holds in the inherently uncertain real-world environment. The talk will concentrate on the consequences of this ignorance towards uncertainties and on several uncertainty-handling strategies that we applied within a real-world vehicle routing application. We will present our results comparing several such strategies both in terms of their ability to trade solution quality for risks as well as their computational efficiency. - 9.10.2024
RNDr. Dalibor Klusáček, Ph.D.
How to manage fairness in a distributed computing system
Abstract: Scientific computing centers or private (in-house) cloud data centers do not rely on the standard pay-as-you-go business model common in commercial clouds to allocate resources. Instead, the system is typically shared by a set of selected users, and the administrator’s job is to ensure that resources are shared fairly given the existing policies of that organization. One common approach, especially in batch systems, is to deploy a fairshare-prioritized scheduler, where a prioritization mechanism balances resource consumption so that individual users get the right shares of resources over time. This talk presents various methods and tools to maintain user-to-user fairness in a production system. Using a set of experiments, we demonstrate multiple approaches to establish and fine-tune the fair-sharing mechanism in a real distributed system, presenting the impact of often-overlooked additional options for modifying the basic fair-sharing settings. - 16.10.2024
RNDr. Terézia Slanináková, RNDr. Matej Antol, Ph.D.
AlphaFind: discovering structure similarity across all known proteome data
Abstract: The evolution of protein structure databases, from the Protein Data Bank (PDB) in the 1970s to the AlphaFold system in 2021, has led to an unprecedented expansion in known protein structures. Current databases now contain hundreds of millions of predicted structures. This vast increase in data offers immense opportunities for biological research but also presents challenges in terms of accessibility and practical use. To bridge the gap between the expansive AlphaFold Database (containing 214 million proteins and occupying 21 TB of storage) and structural biology researchers, we developed AlphaFind. This web-based similarity search system allows for rapid detection of similar proteins. In contrast to traditional search methods based on metadata, AlphaFind employs a purely data-centric search strategy, extracting semantic information directly from the protein features themselves. In this talk, we will discuss the growing trend of using vector embeddings for complex data representation. We will explore how this approach could contribute to the FAIRification (Findable, Accessible, Interoperable, and Reusable) of data repositories and/or help develop a robust tooling ecosystem built on top of these data repositories. - 23.10.2024
Mgr. et Mgr. Jaroslav Oľha
Data-driven dynamic autotuning: A dissertation thesis
Abstrakt: Modern HPC applications need to be programmed in a hardware-aware manner, which can be quite a challenge in an era of large-scale heterogeneous computing setups. Source-code autotuning provides a solution, allowing for definition of many code implementations ahead of time, and switching between them as necessary based on the actual execution environment. However, the autotuning process itself imposes non-trivial overhead, possibly leading to situations where tuning is actively detrimental to overall run time, as it consumes more resources than can be re-gained by the optimized HPC application. This is a rarely addressed problem in autotuning research, and the main focus of my recently finished dissertation thesis. - 30.10.2024
prof. RNDr. Luděk Matyska, CSc.
EOSC — Why should Czech scientists care?
Abstract: We will present the current state of EOSC implementation in Czechia, focusing on the key projects EOSC-CZ and the National Repository Platform (NRP) for research data. The presentation will highlight specific services, tools, and methodologies these initiatives are expected to provide to the Czech research community. Following an introduction to the Czech EOSC Secretariat and its role in supporting scientists, we will explore FAIR data management, discussing its benefits and how it is and will be supported through EOSC implementation. The presentation will conclude with a brief overview of the aims of the upcoming Open Science II project, which is currently under preparation. - 6.11.2024
Mgr. Zdenka Dudová, Ph.D.
TBA - 13.11.2024
doc. RNDr. Radka Svobodová, Ph.D.
TBA - 20.11.2024
Mgr. Pavel Novák
Transforming weakness into strenght: Improving unreliable malware detection methods
Abstract: TBA
Past seminars
- Program in spring 2024
- Program in autumn 2023
- Program in spring 2023
- Program in autumn 2022
- Program in spring 2022
- Program in autumn 2021
- Program in spring 2021
- Program in autumn 2020
- Program in spring 2020
- Program in autumn 2019
- Program in spring 2019
- Program in autumn 2018
- Program in spring 2018
- Program in autumn 2017
- Program in spring 2017
- Program in autumn 2016
- Older seminars
Contact: Hana Rudová