Normal view

There are new articles available, click to refresh the page.
Before yesterdayIndico Feed [Home]

Antonio Manso: Duality symmetry and anomaly for gravitational waves in curved spacetimes

14 May 2026 at 09:00

The vacuum Einstein equations admit a formulation closely analogous to the source-free Maxwell theory. In particular, the linearized equations exhibit an electric-magnetic duality symmetry. We develop a framework that makes this analogy manifest by explicitly identifying the electric and magnetic components of perturbative gravitational waves. Within this formulation, we show that duality rotations between these gravitoelectric and gravitomagnetic fields constitute a Noether symmetry of the linearized theory, and we derive the associated conserved current. The corresponding conserved charge encodes the difference in intensity between the right- and left-handed circularly polarized components of the gravitational wave—that is, between its self-dual and antiself-dual parts. Remarkably, this conservation law remains valid even when the gravitational perturbations propagate on generic curved backgrounds. We then investigate whether this symmetry survives quantization. While the duality symmetry is preserved at the quantum level in flat spacetime, we find that it is anomalously broken in curved backgrounds. As a result, an imbalance between right- and left-handed gravitons could be excited from the vacuum. This effect represents a chiral anomaly for massless spin-two fields, generalizing known results for fermions and spin-one photon fields.

SLAIF - Umetna inteligenca za javni sektor: priložnosti, orodja in prvi koraki

18 May 2026 at 07:30

Javni sektor stoji pred eno največjih priložnosti zadnjih desetletij: umetna inteligenca lahko korenito spremeni način, kako državne institucije obdelujejo podatke, sprejemajo odločitve, komunicirajo z državljani in izvajajo storitve. Hkrati prinaša izzive — od varstva podatkov in regulatornih zahtev do kompetenc uslužbencev in zaupanja v algoritmične sisteme.

Slovenska tovarna umetne inteligence (SLAIF) je v edinstveni poziciji, da javnemu sektorju ponudi podporo brez omejitev, ki veljajo za trg: dostop do suverene računalniške infrastrukture, varnega podatkovnega okolja, pripravljenih orodij UI in strokovnega usposabljanja. SLAIF je del evropske mreže 19 tovarn umetne inteligence, ki se gradijo s skupno naložbo 2,6 milijarde EUR.

Na Odprtem dnevu za javni sektor bomo predstavili, kaj SLAIF konkretno ponuja institucijam javne uprave — od podpore odločanju in analitike do jezikovnih tehnologij za slovenščino, obdelave prostorskih podatkov in klepetalnih robotov. Prisluhnili bomo tudi kolegom iz MKGP, MZ, FURS in pravosodja, ki bodo delili izkušnje z uvajanjem UI v praksi.

Vabljeni so zaposleni v javni upravi, ministrstvih, agencijah, zavodih in občinah — na vseh ravneh, ne le tehnični. Dogodek je zasnovan tako, da nagovori tako odločevalce kot strokovnjake, ki se srečujejo z izzivi digitalizacije in podatkovne analitike.

 

Zakaj se udeležiti?

  • Spoznajte storitve in infrastrukturo SLAIF, ki so dosegljive javnemu sektorju brez omejitev državnih pomoči
  • Prejmite praktične vpoglede iz ministrstev in agencij, ki UI že uvajajo v vsakodnevno delo
  • Odkrijte možnosti usposabljanj za svoje ekipe, od tehničnih do menedžerskih profilov
  • Vzpostavite neposredne stike z raziskovalci IJS in partnerji SLAIF, ki so pripravljeni na pilotna sodelovanja

O SLAIF

SLAIF (Slovenska tovarna umetne inteligence) je nacionalna tovarna umetne ingeligence (AI Factory), sofinancirana v okviru evropskega programa EuroHPC in programa Digitalna Evropa. Koordinira jo Institut »Jožef Stefan« v konzorciju desetih partnerjev. Cilj SLAIF je zagotoviti slovenskim podjetjem, javnim institucijam in raziskovalcem dostop do vrhunske infrastrukture UI, podatkovnih storitev, pripravljenih modelov UI in strokovne podpore — vse na enem mestu.

Uporaba Arnesove infrastrukture in storitev za raziskovalno delo in inovacije

18 May 2026 at 11:00

Učni cilji:
Poznavanje Arnesovih storitev in infrastrukture, varnega dostopa in poti podpore.

Vsebina:
Delavnica udeležencem ponudi celovit pregled ključnih Arnesovih storitev in infrastrukture, ki podpirajo raziskovalno delo. Namen delavnice je raziskovalcem predstaviti storitve, ki so jim na voljo, kako jih varno in učinkovito uporabljati ter kje poiskati podporo pri načrtovanju in izvajanju raziskovalnih projektov. V uvodu bodo predstavljene storitve s področja identitete in dostopa (AAI) in sodelovalna orodja (Arnes E-pošta, Arnes Planer, Arnes Filesender). Udeleženci bodo spoznali tudi možnost uporabe računskih zmogljivosti Arnesovega superračunalnika in podatkovnih storitev (Arnes Shramba). Predstavljeni bodo tudi varnostni vidiki uporabe Arnesove infrastrukture, tipične poti podpore ob vprašanjih in težavah ter dobre prakse za učinkovito rabo. Manjkali ne bodo niti nasveti, kako lahko raziskovalci svoje rezultate predstavijo javnosti s pomočjo Arnesovih storitev (Arnes Canva, Arnes Video, Arnes Splet). Delavnico je moč prilagoditi tipu udeležencev (raziskovalci, visokošolski učitelji, javni uslužbenci) in predvidenim namenom rabe. 

Učni izidi:
- prepoznati, katere Arnesove storitve so relevantne za posamezen raziskovalni scenarij,
- razumeti korake za varen dostop in upravljanje pravic,
- poznati tipične poti podpore (kam in kako prijaviti potrebe/težave),
- vzpostaviti osnovni kontrolni seznam za projekt (dostopi, podatki, varnost, hramba).

Ciljna publika: Raziskovalci, vodje projektov, IT podpora, knjižničarji/RDM, inovacijski oddelki v javnih ustanovah.

Raven zahtevnosti: Predznanje za udeležbo ni potrebno.

Termin: 18. 5. ob 13.00

Izobraževanje za Akademsko in raziskovalno mrežo Slovenije - Arnes izvaja Lenča Ambrožič.

Odprta koda za raziskovanje

19 May 2026 at 09:00

Učni cilji: 
Poznavanje odprtokodnih licenc, uporaba Git, priprava repozitorijev in znanje citiranja kode ter priprave metapodatkov.

Vsebina: 
Delavnica udeležencem predstavi vlogo odprtokodnih praks v sodobnem raziskovalnem prostoru ter njihov pomen za kakovost, transparentnost in sodelovanje znotraj skupnosti. Delavnica posebno pozornost nameni razumevanju odprtokodnih licenc, upravljanju izvorne kode in zagotavljanju reproducibilnosti raziskovalnih rezultatov. Udeleženci se najprej seznanijo z osnovnimi pojmi odprte kode in odprtokodnih licenc, ki jim sledi obravnava načel FAIR (Findable, Accessible, Interoperable, Reusable). Delavnica v nadaljevanju predstavi praktičen primer uporabe sistema Git za verzioniranje kode. Udeleženci se naučijo, kako strukturirati repozitorij, pripraviti dokumentacijo projekta in učinkovito upravljati izdaje. Udeležencem razloži pravilno citiranje programske opreme in pripravi ustreznih metapodatkov, ki omogočajo prepoznavnost in ponovno uporabo kode. Udeleženci spoznajo pomen odprte kode za pravičen in vključujoč razvoj UI modelov. 

Učni izidi:
- izbrati ustrezno odprtokodno licenco in razumeti z njo povezane osnovne obveznosti,
- uporabljati Git za verzioniranje kode in sodelovanje z drugimi,
- pripraviti reproducibilen in jasno strukturiran repozitorij (README, izdaje (release)),
- pravilno citirati kodo in pripraviti ustrezne metapodatke za objavo.

Ciljna publika: Raziskovalci, mladi raziskovalci, podatkovni analitiki, razvijalci v raziskovalnih skupinah, podporne službe RDM/IT.

Raven zahtevnosti: Predznanje za udeležbo ni potrebno.

Termin: 19. 5. ob 11.00

Izobraževanje za Akademsko in raziskovalno mrežo Slovenije - Arnes izvaja dr. Marko Drobnjak.

AI for Science Course

20 May 2026 at 07:00

The full-day course will introduce AI methods suitable for use in science. It will also present examples of their use in different branches of science, including life sciences, envirnmental sciences and materials science. Finally, it will present the Slovenian AI Factory and the opportunities it offers to scientists via its AI-for-Science vertical.

This course on the topic "AI for Science" is organized by the Slovenian AI Factory (SLAIF). The course will be given by Professor Sašo Džeroski. The lectures will be in English.

In-person or on-line attendance is possible. Registration is mandatory. 

More information on the course

Artificial intelligence is already transforming science across many disciplines, and its future impact is expected to be even greater. Realizing this potential, however, requires addressing challenges specific to scientific work: ensuring that models and their predictions are explainable, learning effectively from the limited labelled data that is typical in science, integrating data with existing domain knowledge, and supporting open and reproducible science through the formalization and sharing of scientific knowledge. This course introduces AI methods developed with precisely these challenges in mind. 

The course covers a range of methods suitable for use in science, including explainable machine learning — with trees and ensembles for multi-target prediction as key examples — that produce accurate yet interpretable (or explainable) models for complex scientific domains. It also addresses learning from limited data through two complementary paradigms: semi-supervised learning, which makes use of unlabelled alongside labelled data, and foundation models, which bring representations learned from vast data to bear on data-scarce problems. Further topics include automated scientific modelling, in which interpretable models of dynamical systems are learned from time series data and domain knowledge, and semantic technologies and ontologies for representing and sharing scientific knowledge. 

The course will also present many examples of applying these methods to problems from different branches of science. The methods will be illustrated with concrete applications in life sciences, environmental sciences, and materials science. The course will conclude with a presentation of the Slovenian AI Factory and the opportunities it offers to the scientific community. 

Attendees will leave with a good overview of the current AI-for-science methodological landscape, a grounding in applications to a variety of sciences and a clear picture of how AI factories (and in particular SLAIF) can support their work in the area of AI for Science. 

Information on the lecturer

Sašo Džeroski is Head of the Department of knowledge technologies at the Jozef Stefan Institute and full professor at the Jozef Stefan International Postgraduate School, both in Ljubljana, Slovenia. He is also a visiting professor at the European Space Agency (Frascati, Italy). He is a fellow of EurAI, the European Association of AI, in recognition of his "Pioneering Work in the field of AI”. He is a member of the Macedonian Academy of Sciences and Arts and a member of Academia Europea. He is past president and current vice-president of SLAIS, the Slovenian Artificial Intelligence Society.

His research interests focus on explainable machine learning, computational scientific discovery, and semantic technologies, all in the context of artificial intelligence for science. His group has developed machine learning methods that learn explainable models from complex data in the presence of domain knowledge: These include methods for multi-target prediction, semi-supervised and relational learning, and learning from data streams, as well as automated modelling of dynamical systems.

Professor Džeroski has lead (as coordinator) many national and international (EU-funded ) projects and has participated in many more. He is currently the coordinator of a large national project titled "Artificial Intelligence for Science". He is also the technical lead of SLAIF, the Slovenian Artificial Intelligence Factory. 

The work of professor Džeroski has been extensively published and is highly cited. It has attracted more than 28580 citations and has an h-index of 77 (in the GoogleScholar database). Prof. Džeroski is the most frequently cited computer scientist in Slovenia (according to the 2025 ranking by Research.com). 

Seminar: Pixi - can it replace containers or is it just another environment manager?

21 May 2026 at 12:00

Description: PIXI can be understood as blurring the line between a package/environment manager (e.g. conda) and lightweight containerization. Like conda, it manages dependencies and reproducible environments at the project level, but it does so with stricter locking, declarative configuration, and cross-platform determinism that more closely resembles container workflows. Unlike full containers (Docker/Singularity), PIXI does not virtualize the OS or filesystem; instead, it focuses on reproducible user-space environments with minimal overhead and faster iteration. In practice, it occupies a middle ground: more reproducible and structured than traditional environment managers, but lighter and less isolated than containers—raising the question of whether strict isolation is always necessary for scientific and development workflows.

Difficulty: Beginner

Date & Time: 21.05.2026  from 14.00 to 15.00

Language: English

Target audience: Data Scientists, Application Developers

Virtual location: ZOOM (only registered participants will see ZOOM link)

Organizer:

Univerza v Ljubljani v leto 2024 ...

Lecturer: 

Name: Luka Leskovec
Description: Scientist and educationalist involved in theoretical physics and supercomputing
E-mail: luka.leskovec@fmf.uni-lj.si

Stakeholder needs to AI Factory use cases: SLAIF user stories for heritable cancer early detection (DISARM–PREDI‑LYNCH–SHIELD)

22 May 2026 at 07:00

Course provider: University of Maribor, Faculty of Electrical Engineering and Computer Science (UM FERI)
Instructors: Izidor Mlakar (UM FERI), Zala Meklav (UM FERI)

Learning objectives: Gain practical knowledge on what FMs exist and where they are useful.

Course content: The course will overview the main methodological ideas on how to build, adapt, and evaluate large self-supervised models for biosignals (e.g., PPG, ECG, EEG …) and motion sensors (accelerometers, gyroscopes) under the messy constraints of real-world wearables (noise, motion artifacts, missingness, device/site shifts, and limited labels). 
The training covers modern foundation-model backbones for time series (CNN and ResNet encoders, Transformers, etc.), pretraining paradigms (masked modeling, contrastive/relative contrastive learning), and representation design choices that matter specifically for physiology (beat-synchronous views, morphology-aware learning). Moreover we will highlight state-of-the-art approaches on merging LLMs and sensor-based models that result in hybrid approaches allowing for human-language interpretation of wearable data (e.g., SensorLM).

Learning outcomes: Practical knowledge on what FMs exist, where they are useful, and how to take them beyond default setups to be useful in their domains and problems.

Ai act, etika in upravljanje umetne inteligence – od strategije do skladnosti z EU AI Act

27 May 2026 at 07:00

Course provider: Technology park Ljubljana (TP LJ) ltd.
Instructors: Aleš Pevc (TP LJ), Matej Kirn (TP LJ)

In the era of regulated artificial intelligence, governance is as critical as technical capability. This workshop will address the regulatory framework of the EU AI Act and guide organizations in aligning innovation with compliance, responsibility, and trust.

The session will provide a structured overview of risk classification, internal governance frameworks, data protection considerations, transparency and explainability requirements, and responsible use of generative AI. Organizational accountability and strategic preparedness for operating in a regulated AI environment will be central themes.

Participants will gain practical insights into establishing an AI governance framework that supports sustainable innovation, mitigates legal and reputational risks, and strengthens stakeholder trust.

 

Learning objective: To build a comprehensive understanding of regulatory requirements and governance frameworks necessary for safe, responsible, and compliant AI deployment.

Key topics: The workshop will examine the EU AI Act, risk-based classification of AI systems, governance structures, transparency obligations, data protection, accountability mechanisms, and responsible AI implementation practices.

Learning outcomes: Participants will be able to assess the compliance status of their AI systems, design a foundational AI governance framework, and implement responsible AI practices aligned with European regulatory standards.

Foundation models for wearable sensor data

27 May 2026 at 07:00

Course provider: Jožef Stefan Institute (JSI)
Instructors: Gašper Slapničar (JSI), Mitja Luštrek (JSI)

Learning objectives: Gain practical knowledge on what FMs exist and where they are useful.

Course content: The course will overview the main methodological ideas on how to build, adapt, and evaluate large self-supervised models for biosignals (e.g., PPG, ECG, EEG …) and motion sensors (accelerometers, gyroscopes) under the messy constraints of real-world wearables (noise, motion artifacts, missingness, device/site shifts, and limited labels). 
The training covers modern foundation-model backbones for time series (CNN and ResNet encoders, Transformers, etc.), pretraining paradigms (masked modeling, contrastive/relative contrastive learning), and representation design choices that matter specifically for physiology (beat-synchronous views, morphology-aware learning). Moreover we will highlight state-of-the-art approaches on merging LLMs and sensor-based models that result in hybrid approaches allowing for human-language interpretation of wearable data (e.g., SensorLM).

Learning outcomes: Practical knowledge on what FMs exist, where they are useful, and how to take them beyond default setups to be useful in their domains and problems.

Proton Therapy Masterclasses 2026

1 June 2026 at 07:00

The Proton Therapy Masterclass is an advanced educational course designed to provide a comprehensive understanding of proton beam therapy, a highly precise form of radiation treatment that allows superior dose control while minimizing exposure to surrounding healthy tissue. The class introduces the physical principles of proton therapy, including the Bragg peak, and explains how these principles translate into clinical advantages over conventional photon-based radiotherapy.

Through expert-led lectures and clinically focused discussions, the masterclass explores real world applications across multiple disease sites, treatment planning strategies, delivery technologies, and quality assurance considerations. Aimed at radiation oncologists, medical physicists, and oncology professionals, the program bridges theory and practice, equipping participants with the knowledge needed to evaluate, implement, and advance proton therapy in modern clinical care.

The students will meet with the leading experts in the field of radiopharmaceutial radiotherapy (RPT) and accelerator design for a new age radiotherapy approaches. The lectures will be followed by a practical session on hadron therapy planning.

The event is organised as part of the 31st International Particle Physics Outreach Group Collaboration meeting in collaboration with Jožef Stefan Institute and University of Ljubljana, Faculty of Mathematics and Physics.

 

Ko umetna inteligenca sreča superračunalnike

4 June 2026 at 07:00

Course provider:  University of Ljubljana, Faculty of Computer and Information Science (UL FRI)
Instructors: Davor Sluga (UL FRI), Uroš Lotrič (UL FRI)

Learning objectives: The course aims to provide participants with:

  • understanding of basic concepts of supercomputing,
  • understanding of the role of supercomputing infrastructure in the development of artificial intelligence,
  • knowledge on how to use supercomputing infrastructure,
  • familiarity with software tools specific to supercomputing systems in the context of developing modern artificial intelligence models,
  • understanding of typical workflows for developing artificial intelligence models.


Course Content: 
The course is intended for beginners who want to learn how supercomputing infrastructure can facilitate and accelerate the development of modern artificial intelligence models. Through practical examples, participants will learn the basic concepts of high-performance computing and how to effectively use supercomputers in the SLING network. There will be a special emphasis on practical work, where participants will adapt the processes of building artificial intelligence models to their own needs and run them on a SLING supercomputing system.

Learning Outcomes: After completing the course, the participant will be able to:

  • explain the basic concepts of supercomputing,
  • determine when and why to use supercomputing resources for artificial intelligence tasks,
  • set up an environment for running jobs related to data preparation, training, and fine-tuning of artificial intelligence models,
  • submit, monitor, and manage jobs for training artificial intelligence models.

Workshop: CFD on HPC – OpenFOAM example

8 June 2026 at 07:00

Description: In the three-day course, the use of the OpenFOAM software package, which is currently the most developed open-source CFD system, will be demonstrated. As the name itself suggests, it is an open-source system that any user can enhance according to their needs. Initially, the use of ParaVIEW, a graphical environment for visually reviewing and processing data from OpenFOAM, will be shown. This will be followed by an explanation of how the OpenFOAM environment, with demonstrations of simple examples. Since the foundation of CFD is the mesh, the use of three open-source mesh generators will be demonstrated: GMSH, BlockMesh, and SnappyHexMesh. Subsequently, the application of various areas within the OpenFOAM environment will be explained and demonstrated, including:

  • Fluid transport
  • Transient simulations
  • Transient data processing (animation, particles in flow)
  • Multiphase flows
  • Multi-region simulation (Multi-region)
  • Running cases in an HPC system utilizing OpenFOAM's parallel capabilities

Difficulty: Advanced

Language: According to applications

Date and time: 8. 6. 2026 from: 9:00 - 13.00 
                           9. 6. 2026 from: 9:00 - 13:00
                          10. 6. 2026 from: 9:00 - 13.00

Max. number of participants: 30

Virtual location: ZOOM

Prerequisite knowledge: The basics of the Linux operating system and the basics of fluid mechanics and Python programming.

Target audience: The training is aimed at students and staff in academia and industry who want to learn more about the OpenFOAM open source CFD platform.

Workflow: The training is on-line, in the mornings. The interactive work is done via remote access to the HPC system at ULFS. 

After the workshop you wil:

  • Be able to connect to HPC@ULFS with NoMachine client and work in HPC Linux environment
  • Understand the theoretical background of the Computational Fluid Mechanics (CFD), especially of the Finite Volume Method (FVM)
  • Be able to set up CFD mesh using different open source programs for CFD mesh design (OF – Block Mesh, GMSH)
  • Be able to setup complete OF case (mesh, pysical model, inital and boundary conditions, ...)
  • Be able to setup and run various OF cases in parallel on an HPC cluster
  • Be able to preview and post-process OF results

 

Organiser:

Lecturers:

Ime: Dr. Aleksander Grm
Opis: Aleksander Grm graduated with a Bachelor's degree in Physics from the Faculty of Mathematics and Physics at the University of Ljubljana. He then completed a Master's degree in Applied Mathematics at ICTP/SISA in Trieste, Italy. After the MSc, he continued his studies at the University of Kaiserslautern in Germany and obtained a PhD in Industrial Mathematics. After the PhD, he worked partly in academia and fully in industry. In 2014, he moved to the University of Ljubljana to work in basic and applied research and to teach young people mechanics and mathematics at the engineering level.
E-mail: aleksander.grm@fs.uni-lj.si 
Ime: Dr. Pavel Tomšič
Opis: He is a research assistant at ULFE and is well qualified for several HPC related topics. He is actively involved in efforts to raise competencies in the field of supercomputing, such as the Partnership for Advanced Computing in Europe (PRACE). He is also coordinator of Erasmus + project SCtrain - a strategic partnership for the transfer of knowledge from supercomputing between Slovenia, Austria, the Czech Republic and Italy. As part of the EuroHPC project for the establishment of European National Competence Centers in the field of supercomputing (EuroCC), he is the champion for Training and Skills Development for NCC Slovenia.
E-naslov: pavel.tomsic@fs.uni-lj.si

FAIR data management for artificial intelligence / Upravljanje podatkov za umetno inteligenco po načelih FAIR

8 June 2026 at 07:00

Course provider: Jožef Stefan Institute (JSI) 
Instructors: Panče Panov (JSI)

Learning objectives: 

  1. Understand AI data assets across the lifecycle: how datasets, labels, dataset splits, features, and evaluation artifacts evolve from collection to reuse;
  2. Apply the FAIR principles to AI work: make data and outputs easier to find, access, combine, and reuse (for teams and future projects);
  3. Create an actionable DMP for AI projects: a lightweight plan that supports reproducibility, handover, and compliance; and
  4. Handle constraints responsibly: recognize sensitive data, ethical considerations, access limitations, and industry vs research expectations.


Course content:

  • AI data assets and data life cycle (raw/processed data, labels, splits, evaluation artefacts);
  • Review of the FAIR principles in the context of AI projects;
  • Finding, accessing and reusing data for AI (including access conditions and licensing);
  • Data interoperability for AI (formats, metadata, label definitions and basic standards);
  • Data repositories and sharing strategies for AI datasets and related artefacts;
  • Dealing with confidential, personal, sensitive and private data, and ethical aspects in AI;
  • Data management plan (AI-focused): structure of an AI-oriented DMP, use of FAIR principles, and examples of best practices and tools;
  • Data management in research and industry: open data/open science vs. industrial constraints (governance, IP, security) in AI projects.


Learning outcomes: By the end of the training, participants can:

  1. Explain the AI data lifecycle and basic good practices (structure, documentation, versioning, provenance);
  2. Perform a basic FAIR check on an AI dataset/project and list concrete “quick wins” (metadata, access statement, license, formats);
  3. Draft a short AI-focused DMP (1–2 pages) that a team can actually follow; and
  4. Identify when extra safeguards are required (personal/confidential data, restricted access, IP) and propose sensible mitigations. 

AI Demo Factory – testbed simulacije za realne poslovne primere

10 June 2026 at 07:00

Course provider: Technology park Ljubljana (TP LJ) ltd.
Instructors: Aleš Pevc (TP LJ), Matej Kirn (TP LJ)

 

The AI Demo Factory will function as a practical validation laboratory where companies test their own data within a secure high-performance computing (HPC) environment and obtain concrete analytical results. The workshop will serve as a bridge between concept and validation — transforming raw data into verified business value.

Participants will simulate real business scenarios, test AI-driven hypotheses, and evaluate the technical and economic feasibility of potential solutions. The session will provide structured guidance from data readiness assessment to interpretation of model outputs in a business context.

The AI Demo Factory will allow companies to experiment safely, validate assumptions, and make data-driven investment decisions without the risks associated with immediate production deployment.

 

Learning objective: To enable companies to validate AI solutions using their own data within a controlled HPC environment and to support informed decision-making regarding further implementation.

Key topics: The workshop will cover data readiness assessment, simulation scenario design, model execution on HPC infrastructure, business interpretation of analytical results, and evaluation of technical feasibility and return on investment.

Learning outcomes: Participants will gain a clear understanding of the technical and business feasibility of AI solutions applied to their own data. They will be able to assess implementation readiness, cost implications, and strategic impact before moving toward production deployment.

CSS_Testing Foundation models for wearable sensor data

15 June 2026 at 07:00

Course provider: Jožef Stefan Institute (JSI) 
Instructors: Panče Panov (JSI)

Learning objectives: 

  1. Understand AI data assets across the lifecycle: how datasets, labels, dataset splits, features, and evaluation artifacts evolve from collection to reuse;
  2. Apply the FAIR principles to AI work: make data and outputs easier to find, access, combine, and reuse (for teams and future projects);
  3. Create an actionable DMP for AI projects: a lightweight plan that supports reproducibility, handover, and compliance; and
  4. Handle constraints responsibly: recognize sensitive data, ethical considerations, access limitations, and industry vs research expectations.


Course content:

  • AI data assets and data life cycle (raw/processed data, labels, splits, evaluation artefacts);
  • Review of the FAIR principles in the context of AI projects;
  • Finding, accessing and reusing data for AI (including access conditions and licensing);
  • Data interoperability for AI (formats, metadata, label definitions and basic standards);
  • Data repositories and sharing strategies for AI datasets and related artefacts;
  • Dealing with confidential, personal, sensitive and private data, and ethical aspects in AI;
  • Data management plan (AI-focused): structure of an AI-oriented DMP, use of FAIR principles, and examples of best practices and tools;
  • Data management in research and industry: open data/open science vs. industrial constraints (governance, IP, security) in AI projects.


Learning outcomes: By the end of the training, participants can:

  1. Explain the AI data lifecycle and basic good practices (structure, documentation, versioning, provenance);
  2. Perform a basic FAIR check on an AI dataset/project and list concrete “quick wins” (metadata, access statement, license, formats);
  3. Draft a short AI-focused DMP (1–2 pages) that a team can actually follow; and
  4. Identify when extra safeguards are required (personal/confidential data, restricted access, IP) and propose sensible mitigations. 

Understanding large language models for science and business

15 June 2026 at 07:00

Course provider: University of Ljubljana, Faculty of Computer and Information Science (UL FRI)
Instructors: Marko Robnik-Šikonja (UL FRI)

Learning objectives: Get acquainted with large language models (LLMs), their architecture and training, leading to their effective use.

Course contents: Large language models are changing the way we write, read, and do intellectual jobs. The lecture presents the working of the transformer architecture of neural networks and focuses on the decoder models, which are used in generative language models, such as ChatGPT. Explaining their construction, pretraining, instruction following, preference alignment, and fine-tuning, the lecture gives the necessary background to understand their behaviour. Based on this, it explains prompting strategies, such as in-context learning and chain-of-thought reasoning. The contents are based on examples from business and science.

Learning outcomes: Knowledge of LLM construction and recommendations for their use.

Workshop: Supercomputing Essentials

17 June 2026 at 08:00

Description: In the course, the participants will familiarize with the architecture of computing clusters, the software stack, and run their first jobs. They will learn to distinguish between login nodes, compute nodes, and data storage systems and will become acquainted with the role of the operating system, the Slurm middleware, and user programs. They will connect to the login nodes, transfer files to and from the supercomputer, execute jobs for video processing, and monitor job execution.

Date and Time: 17/06/2026, 10.00 - 15.00

Language: English

Number of participants: 30

Location: ZOOM (Link available to registered users)

Target audience: Researchers, engineers, students, and anyone who needs additional computational resources for their work.

Acquired knowledge:

  • Understanding the operation and architecture of supercomputers

  • Using the SLURM middleware

  • Basic usage of environment modules and containers

  • Managing files and jobs

  • Basic video processing

 

Organizer:

FRI logo

Lecturers:

Name: Davor Sluga
Web page: https://fri.uni-lj.si/sl/o-fakulteti/osebje/davor-sluga 
E-mail: davor.sluga@fri.uni-lj.si
Name: Ratko Pilipović
Web page: https://www.fri.uni-lj.si/sl/o-fakulteti/osebje/ratko-pilipovic
E-mail: ratko.pilipovic@fri.uni-lj.si

 


Uporabniški modeli in priporočilni sistemi

22 June 2026 at 07:00

Course provider: University of Primorska Faculty of Mathematics, Natural Sciences and Information Technologies (UP FAMNIT)
Instructors: Marko Tkalčič (UP FAMNIT)

Learning objectives: Participants will understand how recommender systems connect user behavior to algorithmic decisions, analyze how backend choices shape user experience, and critically evaluate personalization from technical, ethical, and societal perspectives.

Syllabus:

  • what users experience vs. what systems do,
  • the user model: a digital representation of a person,
  • recommender systems: turning user models into suggestions,
  • from interaction to feedback loop,
  • measuring “good” recommendations,
  • real-world applications: user experience meets infrastructure,
  • trust, transparency, and user control,
  • looking ahead: the future relationship between users and recommenders.

Workshop: The Gray Scott School 2026 @ Slovenia

22 June 2026 at 11:30

Video overview

Overview: The Gray Scott School is an advanced training program dedicated to High-Performance Computing (HPC), led by experts from IJCLab, CNRS, Inria, LUPM, LPNHE & LISN. This summer school, in a unique format and entirely free of charge, is dedicated to programming and optimization on Heterogeneous Architectures.

The school covers the optimisation of computations on different types of hardware (CPU, GPU), presenting their respective characteristics, architectures and bottlenecks. It covers generic optimisation methods applicable to all types of hardware, as well as the various libraries, technologies and languages available to achieve the best possible performance. Ideally, the peak performance of the machine. Through hands-on sessions, lectures, and regular technical webinars, the school equips participants with the skills needed to design, optimize, and scale high-performance applications.

 

How to attend the Gray Scott School 2025:

NCC Slovenia is offering a distance learning in Ljubljana, where one of the various satellites in Europe will take place.

The satellite will take place in hybrid format - the speakers will be present in France, and will stream via Zoom. Our lecturers will be in the room to help participants with access and implementation. Interactive development will be supported by the VEGA supercomputer.

Target audience: The workshop is intended for anyone wanting to learn code parallelization on CPUs and GPUs.

Difficulty: Beginner

Prerequisite knowledge: Basic knowledge of Linux, the Terminal and Python

Skills to be gained: Parallelization, vectorization, CPU and GPU programming 

Max number of participants: 8

Date and location:

  • 22.6 - 2.7.2026 Faculty of Mechanical Engineering, Aškerčeva c. 6, Ljubljana & FMF, Jadranska 21, Ljubljana

 

1. dan: 22.06.2026               ura: od 13:30 do 17:00

2. dan: 23.06.2026               ura: od   9:00 do 18:00

3. dan: 24.06.2026               ura: od   9:00 do 18:00

4. dan: 25.06.2026               ura: od   9:00 do 18:00

5. dan: 26.06.2026               ura: od   9:00 do 18:00

6. dan: 29.06.2026               ura: od   9:00 do 18:00

7. dan: 30.06.2026               ura: od   9:00 do 18:00

8. dan: 01.07.2026               ura: od   9:00 do 18:00

9. dan: 02.07.2026               ura: od   9:00 do 18:00

 

 

Lecturers:

Ime:

Luka Leskovec (luka.leskovec@fmf.uni-lj.si)

Opis:

Znanstvenik ter pedagog, ki se ukvarja s teoretično fiziko in visokozmogljivostnim računalništvom.

Ime:

Pavel Tomšič (pavel.tomsic@fs.uni-lj.si)

Opis:

 

Znanstvenik ter pedagog, ki se ukvarja s strojništvom in visokozmogljivostnim računalništvom.

 

 

 

 


Delavnica: Shranjevanje in objava velikih podatkov v repozitorijih NIOD

19 March 2026 at 09:00

Kratek opis: Delavnica je namenjena predstavitvi postopkov shranjevanja in objave velikih podatkov v repozitorijih NIOD. Udeleženci bodo spoznali, kako učinkovito nalagati velike datoteke ali večje število datotek s pomočjo protokola S3 ter kako te podatke ustrezno opisati z metapodatki in jih objaviti za nadaljnjo uporabo.

Podrobnejši opis: Poseben poudarek bo namenjen pripravi in strukturiranju metapodatkov, ki omogočajo ustrezno opisovanje podatkov ter njihovo kasnejšo najdljivost in ponovno uporabo. Udeleženci bodo spoznali dobre prakse pri opisovanju podatkov ter pomen standardizacije metapodatkov.

V praktičnem delu bodo prikazani konkretni primeri nalaganja podatkov v repozitorij, upravljanja z verzijami ter objave podatkovnih zbirk. Udeleženci bodo pridobili znanja, ki jim omogočajo samostojno delo z repozitoriji NIOD in učinkovito upravljanje večjih količin podatkov. 

Zahtevnost: Napredna

Jezik: Slovenski

Termin: 19. 03. 2026 od 10.00 - 14.00

Omejitev števila udeležencev: 10

Virtualna lokacija: MS TEAMS

Priporočeno predznanje: Osnovno poznavanje dela z ukazno vrstico in osnovni koncepti podatkovnih repozitorijev, velepodatki, S3, repozitoriji, metapodatki, upravljanje podatkov, odprti podatki

Ciljna publika: Raziskovalci, inženirji, študenti, podatkovni znanstveniki, podatkovni analitiki 

Potek izobraževanja: Izobraževanje poteka na daljavo v okolju MS Teams. Udeleženci bodo uporabljali orodja za delo s protokolom S3 ter spletni vmesnik repozitorija NIOD. Praktični primeri bodo vodeni in podprti z demonstracijami.

Na izobraževanju pridobljena znanja:

  • Razumevanje konceptov shranjevanja velikih podatkov v repozitorijih
  • Uporaba S3 protokola za nalaganje podatkov
  • Priprava in upravljanje metapodatkov
  • Objava podatkovnih zbirk v repozitorijih NIOD
  • Dobre prakse pri upravljanju raziskovalnih podatkov

 

Organizator:

Predavatelji:

Ime: Marko Ferme
Opis: Marko Ferme je raziskovalec na Fakulteti za elektrotehniko, računalništvo in informatiko Univerze v Mariboru (UM FERI). Njegova raziskovalna področja so obdelava naravnega jezika, arhitektura porazdeljenih sistemov in visokozmogljivo računalništvo. 
E-mail: marko.ferme@um.si 

 


Workshop: CuPY - calculating on GPUs made easy

19 March 2026 at 12:00

Description: Scientific computing increasingly relies on GPU acceleration to handle large datasets and complex numerical tasks. While traditional CPU-based workflows remain essential, modern research benefits greatly from learning how to harness GPUs in an accessible way through Python. CuPY provides a NumPy-like interface that enables users to offload array computations to the GPU with minimal code changes.

On Day 1, we will cover the motivation for GPU computing, discuss what GPUs are best suited for, and set up a self-contained environment. Participants will learn to use conda/mamba for environment management, install and configure a GPU-ready CuPY setup, and verify its functionality.  

On Day 2, we will focus on the CuPY library itself. We will explore its syntax and functionality, emphasizing similarities and differences with NumPy. Through a series of simple examples, and culminating in a more involved case study, participants will gain the skills to confidently integrate GPU acceleration into their Python workflows.

Difficulty: Beginner

Date & Time:

Day 1: 19. 03. 2026  from 13.00 to 16.00

Day 2: 20. 03. 2026 from 13.00 to 16.00

Language: English

Prerequisite knowledge: Basic knowledge of Linux, the Terminal and some Python

Target audience: The workshop is intended for beginners and others interested in using GPUs with python.

Virtual location: ZOOM (only registered participants will see ZOOM link)

Workflow: The training is live over zoom, in the afternoon. The workshop will combine lecture and practical parts, where your own laptop suffices is needed to gain access to the ARNES gpu cluster.

Skills to be gained:

  • how to setup python on a GPU
  • basics of CuPY
  • a more involved example

 

Max number of participants: /

 

Organizer:

Univerza v Ljubljani v leto 2024 ...

Lecturer: 

Name: Luka Leskovec
Description: Scientist and educationalist involved in theoretical physics and supercomputing
E-mail: luka.leskovec@fmf.uni-lj.si

Delavnica: Shranjevanje in objava velikih podatkov v repozitorijih NIOD

26 March 2026 at 09:00

Kratek opis: Delavnica je namenjena predstavitvi postopkov shranjevanja in objave velikih podatkov v repozitorijih NIOD. Udeleženci bodo spoznali, kako učinkovito nalagati velike datoteke ali večje število datotek s pomočjo protokola S3 ter kako te podatke ustrezno opisati z metapodatki in jih objaviti za nadaljnjo uporabo.

Podrobnejši opis: Poseben poudarek bo namenjen pripravi in strukturiranju metapodatkov, ki omogočajo ustrezno opisovanje podatkov ter njihovo kasnejšo najdljivost in ponovno uporabo. Udeleženci bodo spoznali dobre prakse pri opisovanju podatkov ter pomen standardizacije metapodatkov.

V praktičnem delu bodo prikazani konkretni primeri nalaganja podatkov v repozitorij, upravljanja z verzijami ter objave podatkovnih zbirk. Udeleženci bodo pridobili znanja, ki jim omogočajo samostojno delo z repozitoriji NIOD in učinkovito upravljanje večjih količin podatkov. 

Zahtevnost: Napredna

Jezik: Slovenski

Termin: 26. 03. 2026 od 10.00 - 14.00

Omejitev števila udeležencev: 10

Virtualna lokacija: MS TEAMS

Priporočeno predznanje: Osnovno poznavanje dela z ukazno vrstico in osnovni koncepti podatkovnih repozitorijev, velepodatki, S3, repozitoriji, metapodatki, upravljanje podatkov, odprti podatki

Ciljna publika: Raziskovalci, inženirji, študenti, podatkovni znanstveniki, podatkovni analitiki 

Potek izobraževanja: Izobraževanje poteka na daljavo v okolju MS Teams. Udeleženci bodo uporabljali orodja za delo s protokolom S3 ter spletni vmesnik repozitorija NIOD. Praktični primeri bodo vodeni in podprti z demonstracijami.

Na izobraževanju pridobljena znanja:

  • Razumevanje konceptov shranjevanja velikih podatkov v repozitorijih
  • Uporaba S3 protokola za nalaganje podatkov
  • Priprava in upravljanje metapodatkov
  • Objava podatkovnih zbirk v repozitorijih NIOD
  • Dobre prakse pri upravljanju raziskovalnih podatkov

 

Organizator:

Predavatelji:

Ime: Marko Ferme
Opis: Marko Ferme je raziskovalec na Fakulteti za elektrotehniko, računalništvo in informatiko Univerze v Mariboru (UM FERI). Njegova raziskovalna področja so obdelava naravnega jezika, arhitektura porazdeljenih sistemov in visokozmogljivo računalništvo. 
E-mail: marko.ferme@um.si 

 


Delavnica: Vsebniki na superračunalnikih

15 April 2026 at 08:00

Opis: Raziskovalci se pogosto spopadajo z velikimi računskimi izzivi, na primer pri analizi velikih podatkov, fizikalnih simulacijah, računski kemiji, računski biologiji, napovedovanju vremena, simulacijah dinamike tekočin ipd. Za reševanje mnogih problemov je pogosto na voljo ustrezna programska oprema, ki pa jo je potrebno prilagoditi za izvajanje na izbranem superračunalniku.

Na delavnici si bomo ogledali več načinov nalaganja programske opreme: v domačo mapo, preko okoljskih modulov in vsebnikov. Spoznali se bomo s konceptom virtualnih strojev in vsebnikov ter osvetlili razlike med zasnovo vsebnikov Docker in Apptainer. Naučili se bomo uporabiti že pripravljene vsebnike in na praktičnih primerih spoznali, kako zgraditi enostaven vsebnik Apptainer ter ga zagnati v superračunalniškem okolju. V nadaljevanju si bomo ogledali, kako v vsebnik vključiti podporo za grafične pospeševalnike in procesiranje na več vozliščih.

Delavnica bo praktično usmerjena, vaje bomo izvajali na modernem sistemu HPC.

Zahtevnost: Napredna

Jezik: Slovenski

Termin: 15. 04. 2026 od 10:00 - 15:00

Omejitev števila udeležencev: 30

Virtualna lokacija: ZOOM (povezava bo na voljo samo registriranim udeležencem)

Ciljna publika: raziskovalci, inženirji, študenti, vsi ki potrebujejo več računskih virov pri svojem delu

Priporočeno predznanje: 

  • opravljena delavnica Osnove superračunalništva,
  • razumevanje zgradbe računalniške gruče,
  • delo preko odjemalca SSH (ukazna vrstica, prenašanje datotek),
  • osnovno poznavanje vmesne programske opreme Slurm,
  • osnovno znanje operacijskega sistema Linux in lupine Bash
  • osnovno poznavanje programskega jezika Python

 

Na izobraževanju pridobljena znanja:

  • poznavanje vmesne programske opreme Slurm
  • razumevanje okoljskih modulov in vsebnikov
  • uporaba obstoječih vsebnikov Docker in Apptainer
  • gradnja lastnih vsebnikov Apptainer za izvajanje izbranih programov na superračunalniški gruči
  • raba različnih računskih virov v okoljskih modulih in vsebnikih (procesorska jedra, grafični pospeševalniki, vozlišča)

 

Organizator:

FRI logo

Predavatelja:

Ime: Davor Sluga
Opis: https://fri.uni-lj.si/sl/o-fakulteti/osebje/davor-sluga 
E-mail: davor.sluga@fri.uni-lj.si
Ime: Ratko Pilipović
Opis: https://www.fri.uni-lj.si/sl/o-fakulteti/osebje/ratko-pilipovic
E-mail: ratko.pilipovic@fri.uni-lj.si

 


[DBS seminar] Andraž Stožer & Marko Gosak, "From Ca²⁺ Waves to Insulin Pulses: Experimental and Theoretical Perspectives on β-Cell Network Dynamics"

14 May 2026 at 12:15

Pancreatic β-cells within the islets of Langerhans coordinate their activity to generate pulsatile insulin secretion, a process essential for glucose homeostasis and disrupted during the development of diabetes. This seminar will combine experimental physiology and mathematical modelling to examine how multicellular β-cell networks encode, propagate, and translate Ca²⁺ signals into insulin release. In the experimental part, we will present insights from high-frequency confocal Ca²⁺ imaging, showing how β-cells respond heterogeneously to stimulation, how fast Ca²⁺ oscillations and intercellular waves emerge within islets, and how specific cellular subpopulations may contribute to signal initiation and propagation. We will also discuss conserved and altered features of these dynamics in human islets and during diabetes progression. Building on these findings, the modelling part will address why mathematical abstraction is needed to disentangle the mechanisms underlying such complex multiscale behaviour. We will present a phenomenological multicellular model that links structural gap-junction coupling, timescale-specific functional connectivity, and secretion. The model reproduces experimentally observed slow metabolic oscillations, fast bursts, and ultrafast spikes, and shows how interactions between these temporal domains shape pulsatile insulin output. By integrating high-resolution Ca²⁺ imaging with phenomenological modelling, the seminar highlights how interdisciplinary approaches can reveal the principles governing β-cell network function and its disruption in diabetes.

Delavnica: Osnove superračunalništva

27 January 2026 at 09:00

Opis: Na delavnici se bomo seznanili z zgradbo računskih gruč in programsko opremo na njih ter zagnali svoje prve naloge. Naučili se boste razlikovati med prijavnimi vozlišči, računskimi vozlišči, ter sistemi za shranjevanje podatkov. Spoznali boste vlogo operacijskega sistema, vmesne programske opreme Slurm in uporabniških programov. Povezali se boste na prijavna vozlišča, prenašali datoteke na in iz superračunalnika, zaganjali naloge, s katerimi bomo obdelovali video posnetke, in spremljali izvajanje nalog.

Zahtevnost: Osnovna

Jezik: Slovenski

Termin: 27. 01. 2026 od 10.00 - 15.00

Omejitev števila udeležencev: 30

Virtualna lokacija: ZOOM 

Ciljna publika: raziskovalci, inženirji, študenti, vsi ki potrebujejo več računskih virov pri svojem delu

Na izobraževanju pridobljena znanja:

  • Razumevanje delovanja in zgradbe superračunalnikov
  • Uporaba vmesne programske opreme SLURM
  • Osnovna uporaba programskih okolij in vsebnikov
  • Upravljanje z datotekami in poganjanje nalog
  • Osnovna obdelava videoposnetkov

 

Organizator:

FRI logo

Predavatelji:

Ime: Davor Sluga
Opis: https://fri.uni-lj.si/sl/o-fakulteti/osebje/davor-sluga 
E-mail: davor.sluga@fri.uni-lj.si
Ime: Ratko Pilipović
Opis: https://www.fri.uni-lj.si/sl/o-fakulteti/osebje/ratko-pilipovic
E-mail: ratko.pilipovic@fri.uni-lj.si

 


Delavnica: Hitro razvijanje aplikacij z uporabo velikih jezikovnih modelov

29 January 2026 at 09:00

Kratek opis: Ta delavnica ponuja praktičen uvod v razvoj aplikacij z velikimi jezikovnimi modeli (LLM). Napredek v tehnikah in dostopnosti LLM-jev odpira neprimerljive priložnosti za podjetja, da poenostavijo svoje poslovanje, zmanjšajo stroške in povečajo produktivnost. Udeleženci bodo pridobili temeljito razumevanje in praktično znanje o razvoju aplikacij z LLM-ji z raziskovanjem odprtokodnega ekosistema, vključno s prednastavljenimi modeli, ki omogočajo hiter začetek razvoja. Ob koncu delavnice lahko udeleženci pridobijo uradni certifikat NVIDIA Deep Learning Institute.

Podrobnejši opis: Delavnica ponuja celovit uvod v razvoj aplikacij z velikimi jezikovnimi modeli. Spoznali boste, kako so strukturirani veliki jezikovni modeli in kako jih uporabljati. Pregledali bomo arhitekture transformerjev, vmesnike in intuicije ter kako se skalirajo za doseganje najsodobnejših rešitev LLM.

Raziskovali boste specializirane kodirne modele (encoder models) za naloge, kot so semantična analiza, vektorske predstavitve (embeddings), odgovarjanje na vprašanja in klasifikacija brez učenja (zero-shot classification). Prav tako se boste naučili uporabljati dekodirne modele (decoder models) za generiranje zaporedij, kot so programska koda, neomejeni odgovori in pogovori.

Ob koncu tečaja boste znali uporabljati tehnike upravljanja stanja in kompozicije za vodenje LLM-jev k varnim, učinkovitim in natančnim pogovorom, vključno z implementacijo RAG (Retrieval-Augmented Generation) za dostop do zunanjega okolja. Ob koncu delavnice lahko udeleženci pridobijo uradni certifikat Deep Learning Institute pri NVIDIA.

Zahtevnost: Osnovna

Jezik: Slovenski

Opis poteka izobraževanja: Delavnica poteka na daljavo preko brskalnika na oblačni infrastrukturi.

Priporočeno predznanje: Osnovno poznavanje globokega učenja in udobje pri uporabi PyTorch ter prenosnem učenju. Srednje poznavanje Pythona, vključno z objektno orientiranim programiranjem in uporabo knjižnic.

Ciljna publika: Študenti računalništva in informatike, inženirji, raziskovalci, razvijalci ter vsi, ki želijo razumeti in uporabljati velike jezikovne modele v praksi.

Na izobraževanju pridobljena znanja:

  • Poiskati, uporabiti in eksperimentirati z repozitorijem modelov HuggingFace in pripadajočim API-jem 
  • Uporabljati kodirne modele za naloge, kot so semantična analiza, vektorske predstavitve, odgovarjanje na vprašanja in klasifikacija brez učenja
  • Uporabljati dekodirne modele za generiranje zaporedij, kot so programska koda, neomejeni odgovori in pogovori
  • Uporabljati tehnike upravljanja stanja in kompozicije za vodenje LLM-jev k varnim, učinkovitim in natančnim pogovorom

 

Omejitev števila udeležencev: 30

Virtualna lokacija: MS Teams

Organizator: UM FERI, NVIDIA

 

 

Predavatelja:

Ime: Domen Verber
Opis: Domen Verber je docent na Fakulteti za elektrotehniko in računalništvo Univerze v Mari-boru (UM FERI) ter deluje kot strokovnjak na področju visokozmogljivega računalništva (HPC). Njegovo raziskovalno delo je osredotočeno na visokozmogljivo računalništvo in umetno inteligenco, s čimer se kontinuirano ukvarja že več kot 25 let. Med drugim opravlja funkcijo ambasadorja NVIDIA Deep Learning Institute na UM.
  domen.verber@um.si, deep.learning@um.si

 

Ime: Jani Dugonik 
Opis: Jani Dugonik je raziskovalec na Fakulteti za elektrotehniko, računalništvo in informatiko Univerze v Mariboru (UM FERI). Ukvarja se z raziskavami na področjih obdelave naravnega jezika, evolucijskih algoritmov in umetne inteligence.
  jani.dugonik@um.si

GPT-like transformer model for silicon tracking detector simulation

29 January 2026 at 09:30

Simulating physics processes and detector responses is essential in high energy physics and represents significant computing costs. Generative machine learning has been demonstrated to be potentially powerful in accelerating simulations, outperforming traditional fast simulation methods. The efforts have focused primarily on calorimeters.

This seminar presents the very first studies on using neural networks for silicon tracking detectors simulation. The GPT-like transformer architecture is determined to be optimal for this task and applied in a fully generative way, ensuring full correlations between individual hits. Taking parallels from text generation, hits are represented as a flat sequence of feature values.

The tracking performance, evaluated on the Open Data Detector, is presented for single muons, electrons and pions. Benchmarking is performed on recent generations of GPUs to quantify the computing costs of such simulation setup.

RNA Salon #3

18 February 2026 at 14:00

In the third meeting of the Ljubljana RNA Salon, we will have three short talks (15+5 minutes) given by members of participating groups, followed by informal discussion and exchange of ideas.

The RNA Salon will take place at the National Institute of Chemistry.

Kristjan Muursepp: Baryogenesis from bubble collisions

12 December 2025 at 10:00

In this talk I will consider models connecting baryogenesis with first order phase transition. In particular, I will elaborate on scenarios where the required amount of CP-violation originates from the production of heavy particles after scalar shells interpolating between true and false vacuum collide. I will focus on the case when both heavy particles are produced on-shell as well as the opposite regime when one of the heavy particles is produced off-shell, subsequently decaying to light Standard Model states. I will elaborate on the phenomenological implications of both scenarios.

Workshop on BioExcel Software Tools

11 February 2026 at 07:30

This 1.5-day organized workshop within Adriatic Edition of Bioexcel offers a hands-on introduction to modern computational tools and workflows in biomolecular simulations. Participants (primarily from Italy, Croatia and Slovenia) will explore key software frameworks such as BioExcel Building Blocks (BioBB) for setting up reproducible molecular dynamics workflows, and gain practical experience with protein MD setup and automatic ligand parameterization. Further two topics of the workshop cover integrative modelling with HADDOCK, including a tutorial on antibody–antigen complex modelling and alchemical free energy calculations, followed by a PMX-based practical on ligand modifications.

 

Through a combination of lectures and guided tutorials, the workshop provides both theoretical grounding and practical skills best suited for PhD students and young researchers in computational biology and molecular modelling. 

 

Prerequisites

  • working knowledge of Linux,
  • basic familiarity with molecular modeling software.

 

Date: 11-12 February 2026

Location: Ljubljana, Slovenia

Venue: National Institute of Chemistry, Ljubljana, Slovenia

Room: Grand Lecture Hall, National Institute of Chemistry (Map)

Format: On-site

Fee: 70€ (to be paid after application is accepted; fee includes workshop participation, three coffee breaks and two lunches)

Topics: see Contribution list

 

Participation: 40 applicants will be selected for on-site participation. Applicants from Croatia, Italy and Slovenia will be prioritized for on-site participation. We can only accept participants from EuroHPC Joint Undertaking member institutions. In our selection, we will account for geographical and gender distribution. A letter of motivation needs to be handed in when registering (see form). Participants will need to bring a laptop for participating for the tutorial session and refresh their Linux skills. We recommend reviewing this introduction to the UNIX shell.

 

Registration starts:  5 November 2025

Registration extended to: 20 December 2025

Acceptance: 22 December 2025

Payment deadline: 10 February 2026

 

 

 

 

❌
❌