- informatik-vortraege - lists.rwth-aachen.de

[Informatik-assistenten] Einladung Informatik-Oberseminar Lukas Netz
by Netz, Lukas 26 Sep '24

26 Sep '24

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 07.10.2024, 15:00 Uhr Ort: Informatikzentrum der RWTH Aachen University, Gebäude E3, 2. Etage, Raum 9222 Referent: Lukas Netz, M.Sc. Lehrstuhl für Software Engineering Thema: Model-Driven Method for the Development of Full-Size Web-Based Information Systems Abstract: The implementation of information systems poses a variety of challenges for developers and requires a large number of experts from different application domains. This work presents a methodology in the area of generative model-driven development (MDD) for web-based information systems. The methodology consists of three basic building blocks: LLM4CD, CD2GUI and MontiGem. The first building block LLM4CD is based on generative artificial intelligence (LLMs) and transforms natural language requirements and domain descriptions into class diagrams in CD4A notation. The second building block CD2GUI uses these class diagrams and derives models for user interfaces in GUI-DSL notation. The last building block is MontiGem. It uses both types of models to generate a data-centric information system. An important part of this work is the adaptability of the generated artifacts, this mainly concerns the possibility for modelers to adapt the models generated by the AI and CD2GUI, and to adapt and extend the code generated by MontiGem without having to edit the generated code or models directly. This makes iterative work with the methodology possible without discarding adapted code or adapted models. It is precisely this adaptability that facilitates the transition from an underspecified modeled software prototype to a full-size model-driven information system that can be actively used in industry. In addition to the three building blocks, this thesis also presents the domain-specific modeling languages GUI-DSL and CD4A. The methodology is evaluated and discussed both in parts and as a whole. The practical applicability of the methodology is demonstrated using the case study MaCoCo, an information system used by RWTH Aachen University, which serves as a robust benchmark for the validation of the methodology. Es laden ein: die Dozentinnen und Dozenten der Informatik _______________________________________________ Informatik-assistenten mailing list -- informatik-assistenten(a)lists.rwth-aachen.de<mailto:informatik-assistenten@lists.rwth-aachen.de> To unsubscribe send an email to informatik-assistenten-leave(a)lists.rwth-aachen.de<mailto:informatik-assistenten-leave@lists.rwth-aachen.de> https://lists.rwth-aachen.de/postorius/lists/informatik-assistenten.lists.r…

1 0

Einladung: Informatik-Oberseminar Jan Böker
by Jan Böker 17 Sep '24

17 Sep '24

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 27. September 2024, 10:00 Uhr Ort: Raum 9222 (Informatikzentrum E3, Ahornstraße 55) Referent: Jan Wilhelm Böker, M.Sc. Lehrstuhl für Logik und Theorie diskreter Systeme (Informatik 7) Thema: Structural Similarity of Graphs: Graphons and Homomorphism Densities Abstract: The abundance of graph-structured data in domains like bio-informatics has led to the popularity of both graph embeddings and graph neural networks as techniques for machine learning on graphs. Advancements in this field, however, are mostly driven by intuition and empiricism, and there is a lack of a theoretical foundation that explains both the success and the limits of these techniques. The use of tree homomorphism counts is appealing in order to define graph embeddings that not only are well-founded theoretically but also efficiently computable. Graphons emerged as the limit objects in the theory of dense graph limits and are a generalization of graphs to measurable functions. We develop a framework for graphon similarity based on iterated degree measures, which were recently introduced by Grebík and Rocha in a generalization of Color Refinement, a popular heuristic algorithm for graph isomorphism testing, to graphons. We show that, besides tree homomorphism densities, also message-passing graph neural networks integrate into this framework. We integrate a metric variant of Color Refinement into our framework. Furthermore, we introduce the tree cut distance, a variant of the cut distance from the theory of dense graph limits. Our framework allows us to prove that these four concepts all yield the same notion of similarity of graphons. In the special case of graphs, our framework gives us four similarity measures that are not only equivalent but can also be computed efficiently. We also briefly study the computational complexity of the homomorphism reconstructability problem, which asks whether a given vector of natural numbers can actually be realized as homomorphism counts to a graph. We show that this problem is NP-hard even in a severely restricted form. The k-dimensional Weisfeiler-Leman algorithm is a generalization of Color Refinement to higher dimensions. To initiate the development of a higher-dimensional variant of our framework for graphon similarity, we generalize this algorithm and various of its characterizations to graphons and prove their equivalence. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung: Informatik-Oberseminar Johannes Theissen-Lipp geb. Lipp
by Theissen-Lipp, Johannes 13 Sep '24

13 Sep '24

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Mittwoch, 25. September 2024, 11.00 Uhr Ort: COMSYS E3/ Raum 9222, Ahornstraße 55, 52074 Aachen (2359/222, 2.OG) Referent: Johannes Theissen-Lipp M.Sc. Computer Science 5 – Information Systems and Databases Thema: Semantic Foundations of Dataspaces Abstract: Digital transformation is rapidly reshaping industries, organizations, and societies around the world, driven by the exponential growth of data from the Internet and the Web. This data, encompassing both open and closed resources, drives innovation and provides competitive advantages. However, there is a growing demand for data sovereignty, where individuals and organizations control their data throughout its lifecycle. Initiatives such as dataspaces, funded with €4-6 billion in Europe, and the Internet of Production focus on data autonomy and integration, but interoperability and common understanding remain critical challenges. The fields of Semantic Web technologies and FAIR data provide frameworks for addressing these challenges. These semantic technologies enable seamless data exchange and meaningful interpretation to maximize the potential of data, yet they are not sufficiently utilized in dataspaces. This thesis explores the semantic foundations of dataspaces, aiming to integrate semantic technologies to improve data interoperability and facilitate common understanding among stakeholders. It also advances semantic technologies to enhance their impact. This work proposes new concepts, principles, best practices, and solutions to improve data management in dataspaces. The key contributions of this thesis include: - A Tailored Data Lifecycle Model for Dataspaces: This model addresses the unique challenges and requirements of dataspaces, ensuring a streamlined and structured approach to managing data and metadata throughout their lifecycle. - Principles of Information Models: These models define a common core in dataspaces, providing a structured representation of data, services, participants, and interactions, and serving as a foundation for data sovereignty and interoperability mechanisms. - Extended Data Access Principles: These principles cover various types of heterogeneity, improving data accessibility by managing resource constraints, handling dynamic data changes, and providing rich metadata to increase the quantity and quality of information in dataspaces. - Methods for Improving Common Domain Understanding: This includes techniques for identification, best practices for vocabulary development, involvement of non-experts, effective recommendation systems, evolution of semantic information over time, and improved validation techniques to enhance data interoperability and (re)use in dataspaces. By combining these contributions, this thesis confidently integrates semantic technologies into dataspaces. This integration enhances data and service interoperability and facilitates common understanding within these environments, promoting proper data management and reuse throughout the data lifecycle. Our work provides valuable insights into the effective management of information within the digital ecosystem of dataspaces, contributing to the advancement of knowledge in today's data-driven digital transformation. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 1

Einladung: Informatik-Oberseminar Marco Lutz geb. Grochowski
by Grochowski, Marco 02 Sep '24

02 Sep '24

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Dienstag, 10. September 2024, 10.00 Uhr Ort: Im Süsterfeld 9, 52072 Aachen, Raum 259 Referent: Marco Lutz geb. Grochowski M.Sc. Informatik 11 - Embedded Software Thema: Test Suite Generation and Augmentation for Reconfigurable Industrial Control Software in the Internet of Production Abstract: With the advent of Industry 4.0 and the digitally networked factory, cyber-physical production systems (CPPSs) are reconfigured frequently along their life cycle to adapt to changing customer requirements or market demands. Such reconfigurations are not limited to the hardware but also affect the software of the programmable logic controllers (PLCs) driving these plants. While verification and testing are two techniques capable of alleviating the risk of introducing errors in production code, it is no longer sufficient to rely only on the results obtained by these methods during the commissioning of the CPPS. Even minor incremental reconfigurations to the PLC’s software during the operational phase of the life cycle may introduce regressions that can be quickly overlooked by a developer and therefore need to be reverified. The goal of this thesis is to provide a “push button” analysis for generating test cases after a static reconfiguration. The generated test cases can be injected and monitored during maintenance or virtual commissioning to observe the impact of reconfiguration on the CPPS by the developer. In order to reduce redundancy in test suite generation (TSG) after a structural reconfiguration to the PLC software, symbolic summaries of specific parts of the program should be cached and reused to benefit subsequent analysis. While automatic TSG is an established technique used to generate test suites adhering to structural coverage metrics of PLC software, the generated test suite might not anymore be adequate enough with regards to the coverage metric to ensure the absence of regressions. An indispensable part of regression testing (RT) is test suite augmentation (TSA), which guides the TSG toward the reconfigured behavior and increases the chances of deriving difference-revealing test cases which expose behavioral differences between the program and its reconfigured version. The derivation of new test cases is required to uncover potential regressions after a reconfiguration. To this end, the contributions of this thesis include * heuristics for the scalability of the existing TSG for PLC software, * the reuse of symbolic summaries during TSG of reconfigured PLC software, * and the concept of executing the old and new version of a reconfigured PLC software in one unified program version during TSA. These contributions are evaluated on selected domain-specific benchmarks of varying difficulty, such as the PLCopen Safety suite and the Pick and Place Unit (PPU). Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung zum Informatik-Kolloquium von Prof. Dr. Paula Herber
by Völker, Marcus 30 Aug '24

30 Aug '24

+********************************************************************** * * * 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 0

Einladung Informatik-Oberseminar Yingbo Gao, 28. August 2024
by Stephanie Jansen 22 Aug '24

22 Aug '24

********************************************************************** * * * 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

1 0

Einladung: Informatik-Oberseminar Matthias Naaf
by Matthias Naaf 19 Aug '24

19 Aug '24

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 30. August 2024, 11:00 Uhr Ort: Raum 9222 (Informatikzentrum E3, Ahornstraße 55) Referent: Matthias Naaf, M.Sc. LuFG Mathematische Grundlagen der Informatik Thema: Logic, Semirings, and Fixed Points Abstract: Why does my SQL query return a particular answer? How does Hanna win an infinite graph game against Simon? These questions can be answered with semiring semantics, which was introduced by Green, Karvounarakis, and Tannen (2007) for database provenance and has since been extended to various logics within and beyond the database setting. In this talk, we study semiring semantics for the fixed-point logic LFP, which (despite its name) features both least and greatest fixed points. The common theme in semiring semantics is the interplay of logic and algebra, such as the compatibility with semiring homomorphisms and the use of freely generated semirings to represent provenance information. This becomes particularly interesting for fixed-point logic, where additional algebraic and order-theoretic assumptions on the semirings must be made to ensure the existence of fixed points. A key question is the algorithmic computation of fixed points in these semirings. Several techniques have been developed for least fixed points, with the Newton iteration (Esparza, Kiefer, and Luttenberger, 2010) emerging as the most general technique, but it was not known how to compute greatest fixed points. We address this issue by providing an efficient closed-form solution for greatest fixed points of polynomial systems over absorptive semirings. Using the example of Büchi games, we show how semiring semantics of LFP can be applied to analyse winning strategies in infinite games. Lastly, we take the perspective of model theory and study how the classical 0-1 laws about the asymptotic behaviour of logic on random structures can be generalised to semiring semantics of first-order logic. By combining our proof with the fixed-point computation, we further extend the 0-1 law to semiring semantics of LFP. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung Informatik-Oberseminar Eva Fluck
by Fluck, Eva 05 Aug '24

05 Aug '24

+********************************************************************** * * * 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 0

Einladung: Informatik-Oberseminar Malte Breuer
by Malte Breuer 01 Aug '24

01 Aug '24

+********************************************************************** * * * 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 0

Einladung: Informatik-Oberseminar Hinrikus Wolf
by Cormann, Silke 24 Jul '24

24 Jul '24

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Donnerstag, 08. August 2024, 10:00 - 11:00 Uhr Ort: Raum 9222 (großer Seminarraum, Erweiterungsbau 3, Informatikzentrum, Ahornstraße 55) Referent: Herr Hinrikus Eike Wolf, M. Sc. Lehrstuhl Logik und Theorie diskreter Systeme (Informatik 7) Thema: Learning on Graphs from Theory to Industrial Application in Power Management of Distribution Grids Abstract: Learning on graphs has strong ties to theoretical computer science, as some algorithms used for learning are rooted in graph theory. Furthermore, expressivity of learning methods is analysed with techniques from theoretical computer science. From a practical perspective, graph learning finds application in a wide range of domains, such as biochemistry, social science, and in case of this thesis in power management of electric grids. An illustrative example for graph learning is to predict whether a chemical molecule is toxic or non-toxic. The task behind this example involves predicting properties of the whole graph. Beyond this, graph learning includes also to node level tasks, and link prediction. We propose structural node embeddings motivated from Lovasz' (1967) theory of graph homomorphism counts. These are the number of mappings from Graph H to G such that vertices which are adjacent in H are also adjacent in G. The node embeddings consist of vectors representing homomorphism counts from families of graphs within the graph to be embedded. We showcase that our approach achieves comparable accuracy to other methods on benchmark data, except for recent GNN architectures. We conduct a study of the stability of node embeddings across five prominent methods. Most embedding techniques inherently depend on randomness. We analyse the effects of this randomness on the embeddings themselves and on downstream tasks, uncovering significant instabilities, particularly in individual predictions. This finding is crucial for practitioners in selecting an embedding method that meets the requirements of their tasks. We present a GNN architecture for the AC Power Flow problem, which helps detect congestion in AC (alternating current) grids. The AC Power Flow is a non-linear, complex optimization problem without a closed-form solution and is typically addressed using Newton's iterative method. Experimentally, we demonstrate that our method is able to generalize to unknown grids. While the model is better than previous neural approaches, it is not accurate enough to replace classical solvers. We introduce a deep reinforcement learning architecture that can resolve congestion detected by AC Power Flow computation. Congestions appear more often in electric grids, due to the increasing number of electric vehicles, heatpumps, and photovoltaic systems. As in contemporary grids measuring infrastructure is only sparsely available, our architecture learns from this sparse data to resolve the congestions. We demonstrate the ability of our method by experiments on a real-world low voltage grid. Our approach matches accuracy of state-of-the-art classical solvers, with the distinct advantage of being orders of magnitude faster. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0