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

Einladung: Informatik-Oberseminar Tim Quatmann
by Tim Quatmann 23 Aug '23

23 Aug '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Dienstag, 5. September 2023, 14.00 Uhr Ort: Informatikzentrum, E3, 2. Etage, Raum 9222 Referent: Tim Quatmann M.Sc. Lehrstuhl Informatik 2 (Software Modeling and Verification) Thema: Verification of Multi-Objective Markov Models Abstract: Probabilistic systems evolve based on environmental events that occur with a certain probability. For such systems to perform well, we are often interested in multiple objectives, i.e., quantitative performance measures like the probability of a failure or the expected time until task completion. Sometimes, these objectives conflict with each other: minimizing the failure probability possibly means completing the task takes longer. Compromises need to be found. We consider Markov models — particularly Markov decision processes (MDPs) and Markov Automata (MAs). These state-based modeling formalisms describe a system in its random environment. Starting from an initial state, the transitioning behavior in MDPs is determined by probabilistic and nondeterministic choices. MAs further extend MDPs by exponentially distributed continuous time delays. Rewards can be attached to states or transitions to model system quantities such as energy consumption, productivity, or monetary costs. Objectives are formally specified by a mapping from (infinite) system executions to the value of interest, i.e., the total accumulated costs or the average energy consumption. The expected value of an objective is defined once the nondeterminism is resolved using a strategy — intuitively reflecting the choices of a system controller. Different strategies induce different expected objective values. Multi-objective verification of MDPs and MAs analyzes the interplay between the considered objectives and identifies which trade-offs between expected objective values are possible, i.e., achievable by some strategy. We study practically efficient methods to compute the set of achievable solutions. For this, we establish a general framework and its instantiation for (undiscounted) total reachability reward objectives, long-run average reward objectives, and reward-bounded objectives. We propagate the errors made by approximative methods, yielding sound under- and over-approximations. We further consider multi-dimensional quantiles that ask under which reward constraints a given objective value is achievable. Finally, we investigate a setting in which the strategies must be simple, i.e., non-randomized and with limited memory access. All presented approaches are integrated into the state-of-the-art probabilistic model checker Storm. An extensive evaluation of this implementation on a broad set of multi-objective benchmarks shows that our approaches scale to large models with millions of states. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung: Informatik-Oberseminar Jan Pennekamp
by Förster, Claudia 14 Aug '23

14 Aug '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 22. September 2023, 11:00 Uhr Ort: Seminarraum Informatik 4 (COMSYS) - 9007, E3, Ahornstr. 55 [1] Digitaler Zugang (hybrider Vortrag): https://rwth.zoom.us/j/63466099726?pwd=eUN1QTRISUZmSEFzUnhnSXdRSW5CZz09 (Meeting-ID: 634 6609 9726 Kenncode: 269470) Referent: Jan Pennekamp M.Sc. Lehrstuhl Informatik 4 (COMSYS) Thema: Secure Collaborations for the Industrial Internet of Things Abstract: The Industrial Internet of Things (IIoT) leads to increasingly-interconnected and networked industrial processes and environments, which, in turn, results in stakeholders collecting a plethora of information. Even though the global sharing of information and industrial collaborations in the IIoT promise significant improvements concerning productivity, sustainability, and product quality, among others, information is still commonly encapsulated locally. Confidentiality concerns are the primary roadblock to fully realizing the aforementioned improvements. We address this mission-critical research gap. Since existing concepts for sharing information do not scale to industry-sized applications in the IIoT, we present solutions that enable secure collaborations in the IIoT while providing technical (confidentiality) guarantees to the involved stakeholders. Our research is crucial (i) for demonstrating the potential and added value of (secure) collaborations and (ii) for convincing cautious stakeholders of the utility and benefits of technical building blocks that reliably enable confidential information sharing, even among direct competitors. Our interdisciplinary research thus focuses on establishing and realizing secure industrial collaborations in the IIoT. We rely on well-established building blocks from private computing (i.e., privacy-preserving computations and confidential computing) to reliably realize them. We thoroughly evaluate each of our designs using multiple real-world use cases from the domain of production technology to attest their practical feasibility for the IIoT. By applying private computing, we are indeed able to reliably secure collaborations that not only scale to industry-sized applications but also allow for use case-specific configurations of confidentiality guarantees. Overall, given the expected improvements, our research should greatly contribute to convincing even cautious stakeholders to participate in (reliably-secured) industrial collaborations. Our work is an essential first step for establishing widespread information sharing between stakeholders in the IIoT, and it stresses four crucial aspects: (i) collaborations can be secured reliably, and we can even provide technical guarantees while doing so, (ii) building blocks from private computing scale to industrial applications and satisfy the outlined confidentiality needs, (iii) improvements that follow from industrial collaborations are within reach, even when dealing with cautious stakeholders, and (iv) the interdisciplinary development of sophisticated yet appropriate designs for use case-driven secure collaborations can succeed in practice. Es laden ein: die Dozentinnen und Dozenten der Informatik [1] https://www.comsys.rwth-aachen.de/fileadmin/misc/how-to-get-to-comsys.pdf -- Jan Pennekamp, M.Sc., Ph.D. Student Chair of Communication and Distributed Systems RWTH Aachen University, Germany tel: +49 241 80 21411 web: https://www.comsys.rwth-aachen.de/team/jan-pennekamp/

1 0

Einladung: Informatik-Oberseminar Julian Miller
by Miller, Julian 07 Aug '23

07 Aug '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 18.08.2023, 09:00 Uhr Ort: IT Center, Kopernikusstraße 6, Seminarraum 001 Referent: Julian Miller M.Sc. Lehrstuhl Informatik 12 Thema: Pattern-based Abstractions for Parallel Programs Abstract: The computational demands in sciences and engineering are quickly rising with the increasing complexities of simulation and the availability of extensive data. This demand is satisfied with large clusters of computers and specialized hardware accelerators. However, programming such systems is challenging and time-consuming, and the massive concurrency is error-prone. The software developers are faced with deriving a well-scaling solution while preserving correctness. These development challenges are aggravated by the quickly evolving hardware landscape of high-performance computing (HPC). This work investigates the key challenges when developing highly productive and performant parallel programs. This analysis is based on extensive human-subject studies with a diverse set of parallel programs and programmers. It uncovers quantitative and measurable productivity metrics, the main impact factors for developing parallel programs efficiently, and cost estimation methods for developing software. Based on this analysis, an abstract model of parallel algorithms is proposed to mitigate these challenges. It is based on a strict separation between the algorithmic structure of a program and its executed functions. Rich and high-level optimization potentials are revealed by decomposing parallel programs into a hierarchical structure of parallel patterns. A static performance model and optimization and scheduling algorithms are introduced to leverage these optimization potentials. A proof of concept development pipeline is proposed exposing this pattern-based programming approach to software developers: First, parallel programs may be specified in the proposed Parallel Pattern Language (PPL) that closely follows the mathematical definition of parallel algorithms. Alternatively, existing codes can be translated into the proposed hierarchical pattern structure with pattern- detection methods. Second, the hierarchical pattern structure is extracted and global transformations are applied to minimize the overall runtime for a target hardware architecture. Third, the optimized code and its scheduling are generated in a source-to-source fashion for heterogeneous systems with shared and distributed memory and accelerators. The proposed approach and proof of concept implementation are evaluated on real-world parallel algorithms. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung: Informatik-Oberseminar Matthias Ehlenz
by Ehlenz, Matthias 07 Aug '23

07 Aug '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 18. August 2023, 14.00 Uhr Der öffentliche Vortrag findet hybrid statt: Ort: Raum 5053.2 (großer B-IT-Hörsaal)/Informatikzentrum, Ahornstraße 55 Zoom: https://rwth.zoom.us/j/67829456459?pwd=RFM0SjNPb2xidzNIT0xJMUg1UkZGQT09 Meeting-ID: 678 2945 6459 Kenncode: 280923 Referent: Dipl.-Gyml. Matthias Ehlenz Lehr- und Forschungsgebiet Informatik 9 (Lerntechnologien) Thema: A Sustainable Research & Development Ecosystem for Computer-Supported Collaborative Learning with Interactive Tabletop Displays Abstract: Computer-supported learning processes are of consistently increasing importance for education. Considering innovative platforms, interactive tabletops provide distinct advantages: They introduce automated feedback, individualization capabilities, and interaction mechanics that enable self-regulated learning in computer-supported collaborative learning processes. They allow truly equal, simultaneous interaction. At the same time, the strengths of face-to-face communication of traditional group work are preserved. Additionally, the digitization of collaborative face-to-face learning opens new research opportunities by enabling the usage of methods from learning analytics and introducing innovative ways of multi-modal data collection. Still, we do not know much about the practical impact of interactive tabletops on the learning process and successful usage in education requires the tuning of content and media to fit each other. Previous promising technologies missed out on their potential by lacking content, research, and sustainable concepts. This dissertation aims to help overcome obstacles and thus enable interactive tabletops to have a positive impact on future education. Assisting developers and educators to implement open-source learning games, providing interdisciplinary research teams with methods and tools to produce highly configurable research prototypes and partake in open science, and enabling teachers to improve their students' learning by providing insights through the open learning analytics infrastructure: The Multi-Touch Learning Game (MTLG) ecosystem is intended to take a holistic approach on facilitating interactive tabletops for better learning experiences. This dissertation provides a systematic approach to the requirements of, research on and educational use of interactive tabletop systems. The MTLG ecosystem presents an integrated research and prototyping framework for collaborative learning with interactive tabletop displays. The components are the MTLG core and toolchain (fundamental building blocks for the rapid creation of capable research prototypes); the MTLG infrastructure (server-side components for connecting sessions across devices, managing users and more); the MTLG research components (supporting experimental setups). Considering scientific sustainability, this dissertation goes beyond technical aspects and slightly beyond the scope of this project. In an interdisciplinary effort a sustainable, science-driven proposal for a learning analytics metadata infrastructure has been developed. The overall evaluation is done in a threefold approach: First, case studies are presented to show research in depth. Second, the broad applicability is shown by presenting learning applications of different scopes and subjects. Third, a technical prototype and a corresponding case study are presented to showcase the interplay of components. Conclusively, this dissertation presents a coherent ecosystem of software, methods, and infrastructure to research collaborative learning processes involving interactive tabletop displays, large multi-touch systems placed horizontally for face-to-face learning in groups. Es laden ein: die Dozentinnen und Dozenten der Informatik Dipl.-Gyml. Matthias Ehlenz Koordination & Konzeption MediaLab Lehrerbildungszentrum der RWTH Aachen Kármánstr. 17-19 52062 Aachen +49 241 80 96 435 Ehlenz(a)lbz.rwth-aachen.de<mailto:Ehlenz@lbz.rwth-aachen.de> www.lbz.rwth-aachen.de<http://www.lbz.rwth-aachen.de/>

1 0

TODAY UnRAVeL Survey Lecture: "Biggest Milestones"
by Kassing, Jan-Christoph 13 Jul '23

13 Jul '23

Dear all, this is a reminder for Joost-Pieter Katoen's talk with the title "Can we meet the deadline? Most probably: yes!" taking place today at 12:30 in the B-IT room 5053.2. Please find the details below --- Abstract --- Continuous-time Markov chains are used in systems biology, classical performance evaluation, reliability engineering, physics, and so on. We study a the following analysis problem for CMTCs: how to compute the probability to reach a certain target state within a given deadline? Phrased more practically: how likely is it that all substrates have turned into products in a catalytic chemical reaction within a week? We will show that this probability can be characterised as a unique solution of a Volterra integral equation system, whose computation can be reduced to transient analysis of a slightly modified CTMC. We will show why this problem is of practical relevance, that it can be efficiently solved on CTMCs with millions of states, and why its natural generalisation to stochastic scheduling problems is hard. This result was published in 1999, received a test-of-time award in 2022, and forms the ingredient of many state-of-the-art CTMC tools. ---------------- Part of the programme of the research training group UnRAVeL is a series of lectures on the topics of UnRAVeL’s research thrusts algorithms and complexity, verification, logic and languages, and their application scenarios. Each lecture is given by one of the researchers involved in UnRAVeL. This years topic is "Biggest Milestones - Research at Its Peak", UnRAVeL professors will present the most important milestone of their respective research. All interested doctoral researchers and master students are invited to attend the UnRAVeL lecture series 2023 and engage in discussions with researchers and doctoral students. Note that this is the last lecture for this year! We are looking forward to seeing you today. Kind regards, Jan-Christoph for the organisation committee

1 0

AI Colloquium on 18. July: talk by prof. Heike Vallery + poster
by Mann, Julia 12 Jul '23

12 Jul '23

Dear AI community, The next Artificial Intelligence Colloquium<https://www.ai.rwth-aachen.de/cms/KI/Das-KI-Center/Aktuelle-Veranstaltungen…> will be a double event: a talk, and a poster session! On Tuesday, July 18th at 16:00, Prof. Heike Vallery (RWTH Aachen) will give a presentation on “Challenges and opportunities of data-driven learning in physical robots for neurorehabilitation”. After the talk (45 minutes), there will be a Q&A session of 15 minutes. After the talk and the Q&A, the event will continue with a poster session, which will highlight the latest advancements in various fields of AI. You will have the change of discussing hot AI topics with researchers in the discipline! Heike Vallery’s talk will be hybrid, and will take place in the Generali Hall of the Super C as well as on Zoom. A free registration<https://vms.zhv.rwth-aachen.de/prod/veranstaltungen/SingleSignOn/Shibboleth…> is required for both in-person and remote participants (please register on our web page<https://www.ai.rwth-aachen.de/go/id/bckstd>). You will receive the link directly in the confirmation email. The poster session will not be live-streamed. More information on the event series on the website<https://www.ai.rwth-aachen.de/cms/KI/Das-KI-Center/Aktuelle-Veranstaltungen…>. Please feel free to forward this information to your research groups or any other interested parties! The organising team is looking forward to welcoming you on Tuesday next week. Best regards, Julia Mann Dr Julia Mann Managing Director ______________________________________________ Center for Artificial Intelligence RWTH Aachen University Mies-van-der-Rohe-Strasse 15 office: 123 DE-52074 Aachen phone: +49-241-80 20757 Germany http://www.ai.rwth-aachen.de<http://www.ai.rwth-aachen.de/> [cid:image001.png@01D9B4DD.8079AC50]

1 0

Einladung Informatik-Oberseminar Dr. Jakob Bossek
by Andrea Gibbels 03 Jul '23

03 Jul '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Dienstag, 4. Juli 2023, 13.00 Uhr Ort: Raum 9222, Gebäude E3, Ahornstr. 55 Referent: Dr. Jakob Bossek Lehrstuhl Informatik 14 Thema: Tailored Evolutionary Operators for the Multi-Objective Spanning Tree Problem Abstract: With this talk I want to introduce myself to the RWTH computer science department. In the first part I will go into detail on paper recently accepted in the evolutionary computation journal (Bossek & Grimme, 2023) entitled “Tailored Evolutionary Operators for the Multi-Objective Spanning Tree Problem”. The second part will give a broader overview of my research foci and projects in the fields of Evolutionary Optimisation and Automated Artificial Intelligence. The Minimum Spanning Tree (MST) problem is the challenge of finding a tree in an edge-weighted graph that maintains connectivity of all nodes and has minimal costs among all such trees. The MST problem is a fundamental combinatorial optimisation problem with countless applications, e.g., in the construction of communication networks, medical imaging, or many other areas that range from logistic via graph drawing to power grid network design. The basic single-objective version of the MST problem can be solved efficiently, i.e., in polynomial time, e.g., by Prim’s algorithm. The multi-objective MST (moMST) version though (i.e., multiple weights per edge) is NP-hard and suffers from intractability. Thus, efficient heuristics are needed to approximate the set of optimal trade-off solutions. Evolutionary Algorithms are randomised search heuristics that are among the most successful when it comes to solving NP-hard multi-objective optimisation problems. I will present recent work on the design of several highly biased sub-graph-based mutation operators for the moMST problem. In a nutshell, these operators replace (un)connected sub-trees of candidate solutions with locally (Pareto-)optimal sub-trees. The latter (biased) step is realised by applying Kruskal’s single-objective MST algorithm to a weighted sum scalarisation of a sub-graph. I will detail some runtime complexity results for the introduced operators and demonstrate results that show that the sub-graph based operators beat baseline algorithms from the literature even with severely restricted computational budget in terms of function evaluations on four different classes of complete graphs. Es laden ein: die Dozentinnen und Dozenten der Informatik -- Andrea Gibbels, M.A. Administrative Assistant Chair for AI Methodology / Lehrstuhl für Methodik der Künstlichen Intelligenz RWTH Aachen University E-mail: secret(a)aim.rwth-aachen.de Phone: +49 241 80 21452 Theaterstr. 35-39 52062 Aachen Germany

1 1

Einladung Informatik-Oberseminar Benedikt Pago
by Benedikt Pago 30 Jun '23

30 Jun '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Mittwoch, 12. Juli 2023, 11:00 Uhr Ort: Informatikzentrum der RWTH Aachen University, Gebäude E3, 2. Etage, Raum 9222 Referent: Benedikt Pago, M.Sc. Mathematische Grundlagen der Informatik Thema: Limitations of Choiceless Computation Abstract: A central open question in finite model theory asks whether there exists a logic that captures the complexity class PTIME. Another way of phrasing this is whether every polynomial time algorithm can be efficiently simulated by a choiceless one. Choiceless computation models operate on finite structures, such as graphs, and can only perform computation steps which are invariant under the symmetries of the structure. This guarantees that for any two isomorphic input structures, the outcome of the computation is the same - a property which is typically required of logics. The quest for a logic for PTIME seeks to develop and analyse logics (i.e. choiceless computation models) of increasing expressive power within PTIME. One of the most prominent logics that has been suggested and not separated from PTIME within more than 20 years is Choiceless Polynomial Time (CPT). Obtaining a better understanding of its expressive power is the aim of this thesis; the focus is on its limitations, since only little is known so far in this regard. We develop new approaches and techniques towards strong CPT lower bounds. As a promising candidate for separating CPT from PTIME, we propose the graph isomorphism problem on Cai-Fürer-Immerman (CFI) graphs over unordered hypercubes. The CFI construction is a well-known tool to prove inexpressibility results for logics. Choiceless polynomial time algorithms are known to exist for certain ordered variants of it but we show that these fail on unordered CFI graphs. Going further, we study a broader class of CPT-algorithms for this problem and prove that lower bounds against certain families of symmetric Boolean circuits imply lower bounds for these algorithms. A first lower bound for this kind of circuits is also presented to demonstrate the feasibility of the approach. It remains open to strengthen it further to get the desired undefinability result for the CFI problem in CPT.  As an alternative route, we establish a connection to an algebraic proof system called the extended polynomial calculus. The power of this and similar proof systems is the subject of active research in the field of proof complexity. We show that suitable proof complexity lower bounds would imply the separation of CPT and PTIME, too. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

TODAY UnRAVeL Survey Lecture: "Biggest Milestones"
by Kassing, Jan-Christoph 29 Jun '23

29 Jun '23

Dear all, this is a reminder for Christopher Morris's talk with the title "Weisfeiler and Leman Go Machine Learning: Expressivity and Generalization of Graph Neural Networks" taking place today at 12:30 in the B-IT room 5053.2. Please find the details below --- Abstract --- Graph-structured data is prevalent across various domains, such as chemo- and bioinformatics, image analysis, and social network analysis. As a result, there has been a surge in the development of machine-learning methods explicitly tailored to graphs. Among these methods, (message-passing) graph neural networks (GNNs) have emerged as the dominant paradigm. Despite their practical success, GNNs' capabilities and limitations are understood to a lesser extent. In this talk, we survey results connecting GNNs' expressive power and generalization abilities to a simple heuristic for the graph isomorphism problem---the Weisfeiler-Leman algorithm. ---------------- Part of the programme of the research training group UnRAVeL is a series of lectures on the topics of UnRAVeL’s research thrusts algorithms and complexity, verification, logic and languages, and their application scenarios. Each lecture is given by one of the researchers involved in UnRAVeL. This years topic is "Biggest Milestones - Research at Its Peak", UnRAVeL professors will present the most important milestone of their respective research. All interested doctoral researchers and master students are invited to attend the UnRAVeL lecture series 2023 and engage in discussions with researchers and doctoral students. Note that next week there is no talk and then we have the last talk of this year's survey lecture from Joost-Pieter Katoen with the title "Can we meet the deadline? Most probably: yes!" We are looking forward to seeing you at the lectures. Kind regards, Jan-Christoph for the organisation committee

1 0

Einladung: Informatik-Oberseminar Jip Spel
by Birgit Willms 26 Jun '23

26 Jun '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 7. Juli 2023, 10:00 Uhr Ort: Informatikzentrum der RWTH Aachen University, Gebäude E3, 2. Etage, Raum 9222 Referentin: Jip Spel, M.Sc. Lehrstuhl Informatik 2 Thema:Monotonicity in Markov Models Abstract: Many systems exhibit probabilistic behavior, such as randomized protocols, communication protocols, or biological systems. Probabilistic model checking is a common way to analyze these type of systems. We often model these systems as Markov models and check them against a specification typically given in a probabilistic extension of LTL or CTL. One of the main goals is to analyze the probability to reach or the expected total reward upon reaching a set of target states. When checking for these properties, we assume probabilities to be fixed constants. In practice, however, these probabilities are often not fixed but bounded to a given interval. Moreover, these probabilities can be dependent on other probabilities. Therefore, we consider parametric Markov models, in which the probabilities are given symbolically by rational functions over parameters. Common problems are the almost-optimal synthesis problem (i.e., finding a parameter instantiation such that the expected total reward is 𝜀-close to the real unknown optimal expected total reward) and the feasibility problem (i.e., finding parameter instantiations that reach the target with at least a given probability or expected total reward). In this talk, we introduce and define monotonicity of parametric Markov models. We observe that many systems are monotonic in one or more parameters, e.g., if we increase the probability of a parameter, the expected total reward also increases. We show how we can algorithmically find monotonicity and use this to, e.g., improve existing techniques like parameter lifting to solve the almost-optimal synthesis problem. We experimentally show that we significantly improve the time to determine almost-optimal synthesis. Furthermore, we consider the well-known gradient descent method to deal with the feasibility problem. We show that our implementation of the gradient descent method (as implemented in the modelchecker STORM) outperforms particle swarm optimization and quadratically-constrained quadratic programming and performs akin to sequential convex programming. Finally, we introduce parametric probabilistic timed automata (pPTAs). We extend probabilistic timed automata (PTAs) to pPTAs, such that the probabilities can be unknown. We show that some techniques to reduce PTAs to MDPs also work for pPTAs. An implementation of this reduction for pPTAs is available in the Modest Toolset. Es laden ein: die Dozentinnen und Dozenten der Informatik -- ________________________________________ Birgit Willms Lehrstuhl Informatik 2 Koordinierung GRK UnRAVeL RWTH Aachen University Ahornstraße 55 D-52074 Aachen Tel.: +49 (0)241 80 21209 E-Mail:willms@cs.rwth-aachen.de www:www.unravel.rwth-aachen.de/

1 0