informatik-vortraege August 2024

informatik-vortraege@lists.rwth-aachen.de

5 participants
5 discussions

Einladung zum Informatik-Kolloquium von Prof. Dr. Paula Herber

by Völker, Marcus

+********************************************************************** * * * Einladung * * * * Informatik-Kolloquium * * * +********************************************************************** Zeit: Dienstag, 10. September 2024, 14.00 Uhr Ort: Raum 9222, Gebäude E3, Ahornstr. 55 Referentin: Prof. Dr. Paula Herber University of Münster Title: Formal Verification of Cyber-physical Systems using Domain-specific Abstractions Abstract: Cyber-physical systems have become ubiquitous in our daily lives, and their complexity continually evolves to unprecedented levels. In addition to their heterogeneity and interaction with a physical environment, we see a tremendous increase in the use of learning to make autonomous decisions in dynamic environments. These developments pose significant challenges for ensuring the safety and reliability of cyber-physical systems. Formal methods have the potential to guarantee crucial safety properties under all circumstances, but are incredibly expensive and severely suffer from scalability issues. In this talk, I will summarize some of our recent efforts towards reusable specification and more scalable verification of cyber-physical systems using domain-specific abstractions. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 week, 1 day

Einladung Informatik-Oberseminar Yingbo Gao, 28. August 2024

by Stephanie Jansen

********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * *********************************************************************** Zeit: Mittwoch, 28 August 2024, 15:30 Uhr Ort: Raum 9222, E3, Informatikzentrum Zoom: https://rwth.zoom-x.de/j/66261562709?pwd=2driMW8j0cRJLiYFWNJRqRp4Lya80Z.1 Meeting-ID: 662 6156 2709 Kenncode: 337848 Referent: Yingbo Gao, M.Sc. (Lehrstuhl Informatik 6) Thema: Language Modeling and Machine Translation: Improvements in Training and Modeling Abstract: Substantial improvements in language modeling and machine translation have been achieved since the wide adoption of artificial neural networks. In this talk, we discuss three directions related to training and modeling in neural language modeling and neural machine translation. First, sampling-based training criteria are investigated in order to speed up the training of neural language models with large vocabularies. Second, label smoothing, input smoothing as well as multi-agent training are studied to improve the generalization of neural machine translation models. Finally, a language modeling approach for machine translation is proposed to simplify the architecture of existing translation models. Es laden ein: die Dozentinnen und Dozenten der Informatik -- Stephanie Jansen Faculty of Mathematics, Computer Science and Natural Sciences Chair of Computer Science 6 ML - Machine Learning and Reasoning RWTH Aachen University Theaterstraße 35-39 D-52062 Aachen Tel: +49 241 80-21601 sek(a)ml.rwth-aachen.de www.hltpr.rwth-aachen.de

2 weeks, 2 days

Einladung: Informatik-Oberseminar Matthias Naaf

by Matthias Naaf

2 weeks, 5 days

Einladung Informatik-Oberseminar Eva Fluck

by Fluck, Eva

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 16.08.2024, 10:00 Uhr Ort: Informatikzentrum der RWTH Aachen University, Gebäude E3, 2. Etage, Raum 9222 Referent: Eva Fluck, M.Sc. Lehrstuhl für Logik und Theorie diskreter Systeme Thema: Graph decompositions. A study on decompositions of graphs and their application to logic and data science Abstract: We study decompositions of graphs and their application to the expressiveness of first-order logic with counting quantifiers and clustering data. Besides the decompositions themselves we also consider graph parameters that arise from these decompositions as well as obstructions to the existence of good decompositions. The graph class T_q^k are all graphs which have a bounded tree-depth decomposition that also bounds the width. Its homomorphism indistinguishability relation enjoys a rich connection to the expressiveness of counting logic, as shown by Dawar, Jakl and Reggio (2021). Building upon ideas by Dvořák (2010), we reprove this connection. Our proof strategy also enables to construct a graph class which characterizes equivalence under guarded counting logic. A graph theoretic analysis of T_q^k allows us separate it from the intersection of TW_k−1 , the class of graphs of tree-width at most k−1, and TD_q the graphs of tree-depth at most q, if q is sufficiently larger than k. This also yields a new characterization of tree-depth via treedecompositions and a characterization of T_q^k via a Cops-and- Robber game. To lift this separation to the respective equivalence relations, we show that this game is monotone, that is Robber can never reach a vertex that was cleared in some earlier round. This also give a novel proof of the monotonicity of the Cops-and-Robber game for tree- width without the use of any dual object, like brambles. Another part of the separation of the equivalence classes is to prove that T_q^k and TD_q are homomorphism distinguishing closed, which was conjectured by Roberson (2022). Tangles are a concept to formalize regions of high connectivity in graphs in an unambiguous way. It can be translated to more abstract connectivity systems, where they describe highly connected regions in the underling universe. Recently they emerged as a new approach to clustering data. We find that the non-trivial clusters resulting from the well known single linkage hierarchical clustering algorithm are exactly the tangles of an appropriately chosen connectivity function on the data points. We generalize this connection to a one-to-one correspondence between connectivity functions that are maximum- submodular and dendograms which are a way to represent the result of an arbitrary hierarchical clustering algorithm. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 month

Einladung: Informatik-Oberseminar Malte Breuer

by Malte Breuer

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 16. August 2024, 10:00 Uhr Ort: UMIC_025 (2165|025), Mies-van-der-Rohe-Str. 15, EG Referent: Malte Breuer M.Sc. Lehr- und Forschungsgebiet IT-Sicherheit Thema: Privacy-Preserving Kidney Exchange Abstract: Chronic kidney disease has become one of the most common causes of natural death in our modern society. The preferred treatment for chronic kidney disease is the transplant of a kidney from a living donor, who is typically a close friend or relative of the patient. An impediment that prevents such a living donation is that the found living donor is sometimes not medically compatible with the patient. Kidney exchange enables a patient to still receive a kidney transplant in such a situation by exchanging the living donor with other patients. Nowadays, many countries have centralized systems that organize kidney exchange, often on a nationwide scale. Hospitals can register their associated pairs of patients and medically incompatible donors with a central platform, which then tries to find potential exchanges among all registered pairs of patients and donors. Such a centralized kidney exchange system, however, harbors severe security risks that make the central platform susceptible to manipulation and corruption. The core issue is that the operator of the platform alone is responsible for the entire computation of the exchanges. This, for example, allows the platform operator to manipulate the computation such that a particular patient is treated with priority. This does not only make the platform operator susceptible to corruption but it also makes the platform a prime target for high impact attacks aimed at manipulating the computation of exchanges. The central platform becomes an even more attractive target for attackers as it stores the sensitive data of many patients and donors. Thus, any attack that leads to a data breach has a direct impact on the privacy of the sensitive data of many individuals. The main research goal of this thesis is to develop an alternative approach for kidney exchange that is resistant to manipulation and corruption, and protects the sensitive data of the involved patients and donors. To this end, we propose the model of a privacy-preserving kidney exchange system that follows a decentralized approach, where the computation of exchanges is distributed among a set of so-called computing peers. This model ensures that a computing peer is neither able to manipulate the computation of the exchanges nor to learn any information on the sensitive data of the involved patients and donors. We achieve this by using a cryptographic technique called secure multi-party computation. This allows a set of parties to compute a functionality on their private inputs such that each party only learns its own input and output and what can be deduced from both. The core contribution of this thesis is then the development of secure multi-party computation protocols that enable the computing peers to efficiently compute the exchanges for a set of patients and their associated medically incompatible donors. We evaluate the run time of all our protocols and show that our most efficient protocol scales for the large numbers of patients and donors that are to be expected in practice. Thereupon, we simulate the use of our most efficient protocols in our model of a privacy-preserving kidney exchange system over time using real-world data. Our simulations show that the number of transplants achieved over time in our privacy-preserving model is comparable to the number of transplants achieved in the model that is implemented by the existing kidney exchange systems that are susceptible to manipulation and corruption. Thus, our model allows for the replacement of most existing kidney exchange systems at a small or sometimes even negligible impact on the number of transplants over time, while significantly increasing the security guarantees compared to the existing systems. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 month, 1 week

2024

2023

2022

2021

2020

2019

2018

informatik-vortraege August 2024