Ein Angebot des JRZ ISIA
Die Reading Group hat den Charakter eines lockeren Fachseminars mit Vorträgen einer Länge zwischen 30 und 45 Minuten und anschließender Diskussion. In freundlicher Atmosphäre und unter weitestgehender Themenfreiheit werden etwa Forschungsergebnisse, wissenschaftliche oder technologische Überblicksvorträge oder die Aufbereitung einer Forschungsfrage präsentiert und diskutiert. Der Fokus liegt im Austausch, in der Diskussion. Organisatorische Fragen, Anmeldungen zu Vorträgen oder Vortragsvorschläge bitte an Stefan Huber richten.
Termine 2026
March, 4th, 2026 14: pm, Dejan Radovanovic | The Student Who Knew Too Much: Representation Leakage in Distilled Smart Meter Models
Knowledge distillation is a widely used technique to compress neural networks by transferring predictive behavior from a teacher model to a student model. Traditionally, the student learns from both ground-truth labels and softened teacher outputs on real input data. In this talk, we explore a more restrictive and potentially privacy-relevant setting inspired by subliminal learning.
We investigate whether task-relevant information can leak from a trained smart meter model through auxiliary outputs that were never explicitly supervised. In our setup, a teacher model is trained on weekly load profiles to predict binary household attributes (e.g., presence of a swimming pool). A student model, however, never sees real data or labels. Instead, it is trained solely to match auxiliary logits of the teacher evaluated on synthetic probe inputs.
Our preliminary results suggest that under certain conditions, such as shared initialization and specific noise distributions, the student can partially recover task performance. However, the effects are highly sensitive to architectural and experimental choices, and our findings are not yet stable enough for publication. We aim to discuss open questions, methodological challenges, and the broader implications for privacy and secure model sharing in energy informatics.
Room: HS 151, 14:00 pm
March, 10th, 2026 13:30 pm, Florian Graf | Persistent Homology in Practice: Scaling to Millions of Points with Flooder
Topological data analysis aims to extract patterns from data that are overlooked by classical feature engineering. Its most important tool, persistent homology, provides a way to quantify the shape of data by detecting structures such as connected components, loops, and voids. However, computing persistent homology on large datasets is often impractical due to its unfavorable scaling with the number of data points. In this talk, we introduce Flood complexes and the open-source Python package [Flooder] (https://plus-rkwitt.github.io/flooder/), which overcome this limitation for 3D point clouds and enable computations for more than a million points within seconds. The talk includes an intuitive introduction to topology and persistent homology; only minimal mathematical background is needed.
Room: HS 110, 13:30 pm
March, 19th, 2026 15:15 pm, Andreas Bilke (Department CT) | Open Educational Resources: The idea of Open Source for teaching and learning
Open Educational Resources (OER) are teaching and learning materials that are freely accessible and openly licensed, allowing reuse, adaptation, and redistribution. This presentation introduces the main ideas of OER and explains how they are closely connected to the ideas and practices of the open source movement, particularly regarding transparency, collaboration, and shared knowledge creation. Furthermore, the talk provides an overview of the current state of OER at the University of Applied Sciences Salzburg, highlighting existing initiatives and institutional support structures.
Room: HS 151, 15:15pm
If you want to join, please send an email to Stefan Huber.