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

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

TODAY UnRAVeL Survey Lecture: "Biggest Milestones"
by Kassing, Jan-Christoph 22 Jun '23

22 Jun '23

Dear all, this is a reminder for Gerhard Lakemeyer's talk with the title "The Situation Calculus as Lingua Franca for Reasoning about Action" taking place today at 12:30 in the B-IT room 5053.2. Please find the details below --- Abstract --- The Situation Calculus is almost as old as the field of Artificial Intelligence, and thanks to the groundbreaking work by the late Ray Reiter it is still one of the most used and best understood logical formalisms for reasoning about action and change. In this lecture I will introduce the modern form of the situation calculus due to Ray Reiter and go over extensions such as probabilistic actions and the action programming language Golog built on top of the situation calculus. ---------------- 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. Next week we have a talk from Christopher Morris with the title "Weisfeiler and Leman Go Machine Learning: Expressivity and Generalization of Graph Neural Networks" We are looking forward to seeing you at the lectures. Kind regards, Jan-Christoph for the organisation committee

1 0

Einladung Informatik-Oberseminar Alper Yegenoglu
by Abigail Morrison 19 Jun '23

19 Jun '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 23. Juni 2023,14.00 Uhr Ort: Kopernikusstrasse 6, 52704 Aachen, Seminarraum 003 Referent: Dipl. Inf. Alper Yegenoglu SDL Neuroscience, Jülich Supercomputing Centre, Forschungszentrum Jülich, Institute of Geometry and Applied Mathematics & Informatik i3, RWTH Aachen Thema: Gradient-Free Optimization of Artificial and Biological Networks using Learning to Learn Abstract: The quest to understand intelligence, learning, decision-making, and memory in humans has been a longstanding endeavor in neuroscience. Our brain comprises networks of neurons and other cells, but it remains unclear how these networks are trained to solve specific tasks. In the field of machine learning and artificial intelligence, neural networks are commonly optimized using gradient descent and backpropagation. However, applying this optimization strategy to biological spiking networks (SNNs) presents challenges due to the binary communication scheme via spikes among neurons. In my work, I propose gradient-free optimization techniques that can be directly applied to both artificial and biological neural networks. I utilize metaheuristics such as genetic algorithms and the ensemble Kalman Filter (EnKF) to optimize network parameters and train them to solve specific tasks. This optimization is integrated into the concept of learning to learn (L2L), which involves a two-loop optimization procedure. In the inner loop, the algorithm or network is trained on a set of tasks, while in the outer loop, the hyperparameters and parameters are optimized. Initially, I apply the EnKF to a convolutional neural network, achieving high accuracy in digit classification. I then extend this optimization approach to a spiking reservoir network within the Python based L2L-framework. Analyzing connection weights and the EnKF's covariance matrix offers insights into the optimization process.I further adapt the optimization by integrating the EnKF into the inner loop and update hyperparameters using a genetic algorithm. This automated approach suggests alternative configurations, in contrast to a manual parameter tuning. Additionally, I present a simulation where SNNs guide an ant colony in food foraging, resulting in emergent self-coordination and self-organization. I employ correlation analysis methods to understand the ants' behavior. Through my work, I demonstrate the application of gradient-free optimization techniques utilizing learning to learn. I highlight their effectiveness in optimizing biological and artificial neural networks. Es laden ein: die Dozentinnen und Dozenten der Informatik -- Prof. Dr. Abigail Morrison IAS-6 / INM-6 / SimLab Neuroscience Jülich Research Center & Computer Science 3 - Software Engineering RWTH Aachen http://www.fz-juelich.de/inm/inm-6 http://www.fz-juelich.de/ias/jsc/slns http://www.se-rwth.de Office: +49 2428 8097504 Fax # : +49 2461 61-9460 Pronouns: she/her ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------ Forschungszentrum Juelich GmbH 52425 Juelich Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Stefan Müller Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Dr. Ir. Pieter Jansens, Prof. Dr. Astrid Lambrecht, Prof. Dr. Frauke Melchior ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------

1 0

Einladung Informatik-Oberseminar Maximilian Kloock
by Kloock, Maximilian 19 Jun '23

19 Jun '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 30. Juni 2023, 16.30 Uhr Ort: Raum 2202 (HBau, 2. Stock), Ahornstr. 55 Referent: Maximilian Kloock, M.Sc. RWTH Informatik 11 - Embedded Software Thema: Synchronization-Based Cooperative Trajectory Planning of Networked Vehicles Abstract: Centralized decision-making for a Networked Control System (NCS) suffers from a high computational burden on the planning agent. Distributed agents, which compute distributed decision-making, increase the computational performance of the NCS. In distributed decision-making, agents consider a subset of all agents in their planning. Due to only local system knowledge of the agents, these local plans are potentially inconsistent with the local plans of other agents. Such an inconsistency may lead to the infeasibility of local plans. This thesis proposes an algorithm that utilizes synchronization methods to achieve consistency of local plans in distributed decision-making. The algorithm enables fast computations due to parallel decision-making and achieves feasible decisions. The parallel and iterative algorithm alternates two steps: local planning and global synchronization. In the local planning step, the agents parallelly compute local plans. Subsequently, consistency of the local plans across agents is achieved using global synchronization. The globally synchronized plans act as reference decisions for the local planning step in the next iteration. In each iteration, the local planning guarantees locally feasible plans, while the synchronization guarantees globally consistent plans. The parallel algorithm converges to globally feasible decisions if the coupling topology is feasible. This thesis introduces requirements for the coupling topology to achieve convergence to globally feasible decisions and proposes a centralized and a distributed method to generate coupling topologies that fulfill these requirements. The centralized coupling topology generation method uses its knowledge about all agents to generate time-invariant coupling topologies in a greedy manner. The distributed coupling topology generation method generates time-variant coupling topologies using game-theory. In order to demonstrate the distributed decision-making algorithm in experiments, this thesis proposes an architecture for networked control systems that achieves deterministic and reproducible experiments, even with non-deterministic computation and communication times. The Cyber-Physical Mobility Lab (CPM Lab), an open-source and remotely accessible platform for networked and autonomous vehicles, implements this architecture. Moreover, the CPM Lab provides a suitable interface for rapid prototyping with seamless simulations and experiments. This thesis demonstrates the complete pipeline from uncoupled agents to globally feasible solutions. The distributed decision-making uses a Model Predictive Control (MPC) framework. The evaluation of this thesis uses the CPM Lab as an evaluation platform to evaluate the Distributed MPC (DMPC) algorithm using the coupling topologies generated from the centralized and distributed generation methods. The evaluation shows that the proposed DMPC algorithm achieves feasible solutions while requiring lower computation times and lower communication demands than Centralized MPC (CMPC). The resulting trajectories show only small deviations from the globally optimal trajectories generated by CMPC. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Talk: Silvia Sellán (Univ. Toronto) on Uncertain Surface Reconstruction
by Andreas Longva 16 Jun '23

16 Jun '23

Dear colleagues and students, On Friday 23.06, Silvia Sellán from the Dynamic Graphics Project at the University of Toronto will be visiting us to hold a talk about her recent cutting-edge research on stochastic surface reconstruction for recovering shapes from 3D point clouds. Silvia has a proven track record of publishing innovative research at the top venues of her field, and we in the Computer Animation group are thrilled to have her visit. Details below! */When/*: Friday 23.06, 14:00-15:00. */Where/*: Seminar Room 118, Building E3, Floor 1 Mies-van-der-Rohe-Straße, 52074 Aachen (same floor as the Computer Graphics group) /*Title*/: Uncertain Surface Reconstruction */Abstract/*: We propose a method to introduce uncertainty to the surface reconstruction problem. Specifically, we introduce a statistical extension of the classic Poisson Surface Reconstruction algorithm for recovering shapes from 3D point clouds. Instead of outputting an implicit function, we represent the reconstructed shape as a modified Gaussian Process, which allows us to conduct statistical queries (e.g., the likelihood of a point in space being on the surface or inside a solid). We show that this perspective improves PSR's integration into the online scanning process, broadens its application realm, and opens the door to other lines of research such as applying task-specific priors. /*Bio*/: Silvia is a fourth year Computer Science PhD student at the University of Toronto. She is advised by Alec Jacobson and working in Computer Graphics and Geometry Processing. She is a Vanier Doctoral Scholar, an Adobe Research Fellow and the winner of the 2021 University of Toronto Arts & Science Dean’s Doctoral Excellence Scholarship. She has interned twice at Adobe Research and twice at the Fields Institute of Mathematics. She is also a founder and organizer of the Toronto Geometry Colloquium and a member of WiGRAPH. She is currently looking to survey potential future postdoc and faculty positions, starting Fall 2024. Looking forward to seeing curious researchers and students alike next Friday! Sincerely, Andreas Longva, Msc. Computer Animation group Visual Computing Institute

1 0

TODAY UnRAVeL Survey Lecture: "Biggest Milestones"
by Kassing, Jan-Christoph 15 Jun '23

15 Jun '23

Dear all, this is a reminder for Britta Peis's talk with the title "Ascending Auctions and Matroids" taking place today at 12:30 in the B-IT room 5053.2. Please find the details below --- Abstract --- We consider matching markets, where an auctioneer is selling a set of indivisible items E to a set of bidders with private valuations v_j. In an ascending auction, the auctioneer decides on an assignment of items to bidders, together with prices p_i for all items i in E, based on the demand indicated by the bidders throughout the auction. In the first part of the talk, we consider computability and monotonicity of buyer-optimal Walrasian equilibria in ascending auctions. A price vector p is called "Walrasian", if there exists an assignment under which every bidder j receives a bundle S of maximum payoff v_j(S)-p(S), and, additionally, all items of positive price are sold. Walrasian prices are called "buyer-optimal" if they minimise the total payment of the bidders among all Walrasian prices. It is well-known that buyer-optimal Walrasian prices exist and can be computed efficiently if the valuations v_j are gross-substitute. Moreover, it is known that an ascending auction, which iteratively raises the prices uniformly on items in a so-called "inclusionswise minimal maximal overdemanded set" terminates with buyer-optimal Walrasian prices, and that these overdemanded sets can be computed efficiently with tools from Discrete Convexity. We will see how to compute the desired overdemanded set in each iteration in a simple combinatorial way with the matroid partition algorithm. Moreover, we will use the obtained structural insights to show that the buyer optimal Walrasian prices are monotone in supply and demand. In the second part of the talk, we consider ascending auctions where the auctioneer is restricted to sell a basis of a matroid to the bidders. It is known that a very natural and simple ascending auction is welfare-maximising and truthful in the sense that the prices coincide with the Vickrey payments. We will provide a simple proof of this result, combined with certain sensitivity results. (Joint work with Katharina Eickhoff, Niklas Rieken, and Laura Vargas Koch). ---------------- 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. Next week we have a talk from Gerhard Lakemeyer with the title "The Situation Calculus as Lingua Franca for Reasoning about Action" We are looking forward to seeing you at the lectures. Kind regards, Jan-Christoph for the organisation committee

1 0