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

Einladung: Informatik-Oberseminar Max Lyon
by Max Lyon 27 Mar '23

27 Mar '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 17. April 2023, 10.00 Uhr Ort: Gebäude E3, Seminarraum 118, Ahornstr. 55 Referent: Max Lyon M.Sc. Lehrstuhl für Informatik 8 Thema: Quad Mesh Layout Structure Optimization Abstract: Quad meshes are very useful in many applications such as modeling and animation in the entertainment industry, simulation in the construction industry, or as intermediate representation to form easily editable spline surfaces. Some of the interesting properties of quad meshes compared to triangular meshes are natural alignment of edges to the surface's principal curvature directions and their ability to form highly anisotropic elements without the need of excessively small angles. Therefore, the computer aided or fully automatic generation of quad meshes has been a research focus for many years and is probably going to remain as such for the foreseeable future. One set of methods that have proven to be particularly well suited for the generation of high quality quad meshes are centered around the creation of an integer-grid map, i.e. a parametrization of the surface into the 2D domain whose parametric integer iso-lines induce a quad mesh structure on the input mesh. These methods offer a high degree of control over the desired sizing and alignment of the resulting quads and generally achieve high element quality. Unfortunately, they disregard the high level structure, i.e. how the irregular vertices of the resulting quad mesh are connected by edge strips. Ideally, these edges form a coarse quad layout, enabling the use of highly efficient multigrid solvers or the manual editing of a small number of spline patches. In this thesis we give an overview of current state-of-the-art quad meshing algorithms which we then improve upon in different aspects: We show how a valid quad mesh can always be extracted for the given layout via careful re-embedding of a so called motorcycle graph combined with patchwise harmonic parametrizations followed by a global optimization. We extend the motorcycle graph construction in order to lift the constraints of layouts having to exactly follow the boundaries of the mesh leading to meshes with less distortion whose quads can be cut off by the boundary at an arbitrary angle. Further, we present a method that is guaranteed to yield a coarse quad layout that adheres to a minimum quality as defined by the user defined maximal allowed deviation of layout arcs from desired direction. For cases where strict adherence to the input field is not required we show how singularities may be moved to better locations or removed entirely to further improve the resulting layout structure. Finally, we discuss how a quad mesh or layout may be refined in order to achieve a locally increased resolution or anisotropy. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung Informatik Oberseminar René Röpke
by Lehr- und Forschungsgebiet Informatik 9 23 Mar '23

23 Mar '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Mittwoch, 12. April 2023, 15.00 Uhr Ort: 9222, E3, Ahornstr. 55 und hybrid via Zoom (https://rwth.zoom.us/j/92145754381?pwd=TmhOOXBDc3RITGl1L0NwdGhLTEVKUT09) Referent: René Röpke M.Sc. Lehr- und Forschungsgebiet Informatik 9 (Lerntechnologien) Thema: Extending Game-Based Anti-Phishing Education using Personalization Abstract: Phishing poses an imminent and wide-ranging threat to Internet users worldwide, in which attackers use methods of deception to lure victims into disclosing information. Recent reports state high numbers of phishing incidents and, so far, technical solutions fail to stop the threat completely. As a complementary approach, user education using anti-phishing learning games has been explored to raise awareness and teach the necessary knowledge and skills to detect and protect against phishing attacks. A common game mechanic used in existing games requires learners to classify URLs as either legitimate or phishing in a binary decision scheme. Here, a problem can occur if learners do not know the service of a given URL and are unable to classify the URL due to a lack of reference. As such, learners may revert to guessing which may weaken the game's potential for practice, since learners cannot relate between correct classifications and the applied knowledge. Furthermore, the possibilities for feedback are limited since the binary decision mechanic does not provide any insights into learners' decision processes and possible misconceptions. In this dissertation, the limitations for feedback as well as the problem with classifying unknown URLs in anti-phishing learning games are addressed as follows: First, a review of existing learning games provides insights into their design and covered learning content. Its results are used in guiding the design and implementation of two new game prototypes. Here, the first game extends the before-mentioned binary decision mechanic and requires learners to sort URLs into one of many categories, depending on which manipulation technique was applied to a distinct part of the URL. The second game requires learners to apply different manipulation techniques and create their own malicious URLs using a puzzle mechanic. Next, the means of personalization for anti-phishing learning games are explored and a personalization pipeline is developed. By considering the learners' familiarity with different services and dynamically creating benign and phishing URLs, the content of anti-phishing learning games can be personalized. To evaluate the new game prototypes as well as the application of the personalization pipeline, two comparative user studies are conducted in a between-group design with pre-, post- and longitudinal testing. In the first user study with 133 participants, both games are evaluated and compared to a baseline implementation. While participants of the new games did not perform significantly better than the control group, results show significant improvements in the participants' performance and confidence between pre- and post-tests for all games, as well as notable differences when classifying URLs of unknown and known services. In the second user study with 49 participants, the personalization pipeline is integrated into one of the games, in order to compare its personalized and non-personalized version. Here, personalization enables the control of service familiarity and allows insights into how URLs of unknown services are handled within the game. While participants of the personalized game did not outperform the participants of its non-personalized version, the evaluation of in-game behavior provides insights into learners' decision processes and possible problems or misconceptions. Furthermore, results of a longitudinal evaluation of all games and versions show that knowledge is retained since the participants perform still significantly better than in the pre-test. In all, this dissertation presents first approaches and research results in the domain of personalized anti-phishing learning games. Future work may entail redesigning anti-phishing learning games to incorporate further means of personalization and to understand how learner characteristics can be utilized in anti-phishing learning games. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung Informatik-Oberseminar Weiyue Wang, 15. März 2023, 14:00 Uhr
by Stephanie Jansen 07 Mar '23

07 Mar '23

********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * *********************************************************************** Zeit: Mittwoch, 15. März 2023, 14:00 Uhr Zoom: https://rwth.zoom.us/j/93749104715?pwd=bEFSR0lsR1o1Tzl2enVMSzM0aTE2Zz09 Meeting-ID: 937 4910 4715 Kenncode: 743829 Referent: Weiyue Wang, M.Sc. Thema: Neural Hidden Markov Model for Machine Translation Abstract: Recently, neural machine translation systems have shown promising performance. One of the key components that almost all modern neural machine translation systems contain is the attention mechanism, which helps an encoder-decoder model attend to specific positions on the source side to produce a translation. However, recent studies have found that using attention weights straight out of the box to align words results in poor alignment quality. This inspires us to introduce an explicit alignment model into the neural architecture in order to improve the alignment and thus also the translation quality of the overall system. To this end, we propose a novel neural hidden Markov model consisting of neural network-based lexicon and alignment models trained jointly with the forward-backward algorithm. Various neural network architectures are used to model the lexicon and the alignment probabilities. We start with feedforward neural networks and apply our first model to re-rank n-best lists generated by phrase-based systems and observe significant improvements. In order to build a monolithic neural hidden Markov model, the more powerful recurrent neural networks are applied to the architecture, and a standalone decoder is implemented. By replacing the attention mechanism with an alignment model, we achieve comparable performance to the baseline attention model while significantly improving the alignment quality. We also apply the state-of-the-art transformer architecture to the neural hidden Markov model and the experimental results show that the transformer-based hidden Markov model outperforms the standard self-attentive transformer model in terms of TER scores. In addition to the work on the neural hidden Markov model, we propose two novel metrics for machine translation evaluation, called CHARACTER and EED. These are easy-to-use and perform promisingly in the annual WMT metrics shared tasks. Es laden ein: die Dozentinnen und Dozenten der Informatik -- Stephanie Jansen Faculty of Mathematics, Computer Science and Natural Sciences HLTPR - Human Language Technology and Pattern Recognition RWTH Aachen University Theaterstraße 35-39 D-52062 Aachen Tel: +49 241 80-21601 sek(a)hltpr.rwth-aachen.de www.hltpr.rwth-aachen.de

1 0

Einladung: Informatik-Oberseminar Herr Arnab Chakrabarti
by Maaßen, Leany 06 Mar '23

06 Mar '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 13.03.2023, 15:00-16:00 Uhr Der öffentliche Vortrag findet hybrid statt: Raum: Raum 5053.2 (großer B-IT-Hörsaal)/Informatikzentrum, Ahornstraße 55 Zoom: https://rwth.zoom.us/j/94314167045?pwd=VWt3TTUrUjQrbTRQcDUzYm8yMVcydz09 Meeting-ID: 943 1416 7045 Kenncode: 239326 Referent: Herr Arnab Chakrabarti, M. Sc. Lehrstuhl Informatik 5 Thema: Exploratory Pipeline for Analysis and Visualization of Large Information Spaces Abstract: Visual exploration of datasets having excessive data points and features leads to over-plotting and visual clutter, resulting in the loss of interpretability. This problem has been termed as "visual noise" in the field of high-dimensional data exploration. Our work in this thesis centers around the question of reducing visual noise and improving data interpretability in the quest to find interesting patterns within high-dimensional datasets. In this thesis, we present algorithms and workflows to tackle the so-called "curse-of-dimensionality". We proposed a visualization ranking model capable of comparing diverse visual structures within a unified framework. Finally, we present a novel visualization recommendation system capable of the process of generating meaningful visualizations that help users to explore and understand their data more efficiently. To evaluate the effectiveness of our solutions in a high-dimensional setting, we present a data exploration pipeline and test it within the Internet of Production (IoP) settings. Es laden ein: die Dozentinnen und Dozenten der Informatik _______________________________ Leany Maaßen RWTH Aachen University Lehrstuhl Informatik 5, LuFG Informatik 5 Prof. Dr. Stefan Decker, Prof. Dr. Matthias Jarke, Prof. Gerhard Lakemeyer Ph.D., JunProf. Dr. Sandra Geisler Ahornstrasse 55 D-52074 Aachen Tel: 0241-80-21509 Fax: 0241-80-22321 E-Mail: maassen(a)dbis.rwth-aachen.de

1 0

Einladung zum Informatik Oberseminar von Herrn Hendrik Simon, M.Sc.
by Hendrik Simon 06 Mar '23

06 Mar '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Donnerstag, 4. Mai 2023, 13.30 Uhr Ort: Seminarraum 2202, Hauptbau, Ahornstr. 55 Referent: Hendrik Simon M.Sc. Lehrstuhl Informatik 11 Thema: Automatic Test Case Generation for PLC Software Abstract: Automatic test case generation for the purpose of bug finding or achieving coverage goals has recently evolved to a scalable technique that is nowadays used to find highly critical security bugs, e. g. by Microsoft. However, in the domain of Programmable Logic Controllers (PLCs), applications of this technique are rare and usually rely on tools and mechanisms that were not initially designed for this domain. In fact, a discussion on how to design such techniques with the peculiarities of PLC software in mind, is missing. At the same time, PLC software is typically used in safety critical environments where software errors pose significant threats to the environment or humans and may additionally result in significant financial losses. Mature automatic testing techniques for the PLC domain would, thus, be highly beneficial to further support software quality in this area. PLC software typically follows a cyclic execution scheme that involves a repeated process of reading input values, executing a (often state machine based) control program that relies on local variables and writing computed values to outputs. Although the cyclic execution resembles only a small change in the execution semantics, the impact on automatic testing techniques is significant. This dissertation provides insights and mechanisms to transfer automatic test case generation into the domain of PLC software. We conduct an in-depth discussion on related approaches and point out strengths and weaknesses in order to provide baseline knowledge that can be utilised in future developments in this field of research. Further, we introduce our own automatic test case generation approaches and exemplify their effectiveness on PLC software. We are able to show that the generation of branch coverage tests can be achieved significantly faster than with existing techniques, rendering our approaches more applicable for larger software. The focus of our techniques lies in the exploitation of state-machine based execution behaviour and the preservation of structural information in Sequential Function Chart. For the latter, our presented algorithm can achieve full coverage in a few seconds for programs that could only partly be covered within an hour by related approaches. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung: Informatik-Oberseminar Steffen van Bergerem
by Steffen van Bergerem 21 Feb '23

21 Feb '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 10. März 2023, 11.00 Uhr Ort: Raum 9222, Gebäude E3, Ahornstr. 55 Zoom: https://rwth.zoom.us/j/95679484581?pwd=Q0N4SUFZQ21jVERZMDAwc2cyQXpzZz09 Referent: Steffen van Bergerem, M.Sc. Lehrstuhl für Informatik 7 Thema: Descriptive Complexity of Learning Abstract: Supervised learning is a field in machine learning that strives to classify data based on labelled training examples. In the Boolean setting, each input is to be assigned to one of two classes, and there are several fruitful machine-learning methods to obtain a classifier. However, different algorithms usually come with different types of classifiers, e.g. decision trees, support-vector machines, or neural networks, and this is cumbersome for a unified study of the intrinsic complexity of learning tasks. This thesis aims at strengthening the theoretical foundations of machine learning in a consistent framework. In the setting due to Grohe and Turán (2004), the inputs for the classification are tuples from a relational structure and the search space for the classifiers consists of logical formulas. The framework separates the definition of the class of potential classifiers (the hypothesis class) from the precise machine-learning algorithm that returns a classifier. This facilitates an information-theoretic analysis of hypothesis classes as well as a study of the computational complexity of learning hypotheses from a specific hypothesis class. As a first step, Grohe and Ritzert (2017) proved that hypotheses definable in first-order logic (FO) can be learned in sublinear time over structures of small degree. We generalise this result to two extensions of FO that provide data-aggregation methods similar to those in commonly used relational database systems. First, we study the extension FOCN of FO with counting quantifiers. Then, we analyse logics that operate on weighted structures, which can model relational databases with numerical values. For that, we introduce the new logic FOWA, which extends FO by weight aggregation. We provide locality results and prove that hypotheses definable in a fragment of the logic can be learned in sublinear time over structures of small degree. To better understand the complexity of machine-learning tasks on richer classes of structures, we then study the parameterised complexity of these problems. On arbitrary relational structures and under common complexity-theoretic assumptions, learning hypotheses definable in pure first-order logic turns out to be intractable. In contrast to this, we show that the problem is fixed-parameter tractable if the structures come from a nowhere dense class. This subsumes numerous classes of sparse graphs. In particular, we obtain fixed-parameter tractability for planar graphs, graphs of bounded treewidth, and classes of graphs excluding a minor. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung: Informatik-Oberseminar Herr Aleksandar Mitrevski
by Maaßen, Leany 19 Jan '23

19 Jan '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 30.01.2023, 13:00-14:00 Uhr Ort: Informatikzentrum, Ahornstraße 55, Raum 5053.2 (B-IT-Hörsaal) Referent: Herr Aleksandar Mitrevski, M. Sc. LuFG Informatik 5 Thema: Skill Generalisation and Experience Acquisition for Predicting and Avoiding Execution Failures Abstract: For performing tasks in their target environments, autonomous robots usually execute and combine skills. Robot skills in general and learning-based skills in particular are usually designed so that flexible skill acquisition is possible, but without an explicit consideration of execution failures, the impact that failure analysis can have on the skill learning process, or the benefits of introspection for effective coexistence with humans. Particularly in human-centered environments, the ability to understand, explain, and appropriately react to failures can affect a robot's trustworthiness and, consequently, its overall acceptability. Thus, in this dissertation, we study the questions of how parameterised skills can be designed so that execution-level decisions are associated with semantic knowledge about the execution process, and how such knowledge can be utilised for avoiding and analysing execution failures. The first major segment of this work is dedicated to developing a representation for skill parameterisation whose objective is to improve the transparency of the skill parameterisation process and enable a semantic analysis of execution failures. We particularly develop a hybrid learning-based representation for parameterising skills, called an execution model, which combines qualitative success preconditions with a function that maps parameters to predicted execution success. The second major part of this work focuses on applications of the execution model representation to address different types of execution failures. We first present a diagnosis algorithm that, given parameters that have resulted in a failure, finds a failure hypothesis by searching for violations of the qualitative model, as well as an experience correction algorithm that uses the found hypothesis to identify parameters that are likely to correct the failure. Furthermore, we present an extension of execution models that allows multiple qualitative execution contexts to be considered so that context-specific execution failures can be avoided. Finally, to enable the avoidance of model generalisation failures, we propose an adaptive ontology-assisted strategy for execution model generalisation between object categories that aims to combine the benefits of model-based and data-driven methods; for this, information about category similarities as encoded in an ontology is integrated with outcomes of model generalisation attempts performed by a robot. The proposed methods are evaluated in multiple experiments performed with a Toyota Human Support Robot. The main contributions of this work include a formalisation of the skill parameterisation problem by considering execution failures as an integral part of the skill design and learning process, a demonstration of how a hybrid representation for parameterising skills can contribute towards improving the introspective properties of robot skills, as well as an extensive evaluation of the proposed methods in various experiments. We believe that this work constitutes a small first step towards more failure-aware robots that are suitable to be used in human-centered environments. Es laden ein: die Dozentinnen und Dozenten der Informatik _______________________________ Leany Maaßen RWTH Aachen University Lehrstuhl Informatik 5, LuFG Informatik 5 Prof. Dr. Stefan Decker, Prof. Dr. Matthias Jarke, Prof. Gerhard Lakemeyer Ph.D., JunProf. Dr. Sandra Geisler Ahornstrasse 55 D-52074 Aachen Tel: 0241-80-21509 Fax: 0241-80-22321 E-Mail: maassen(a)dbis.rwth-aachen.de

1 0

Einladung Informatik-Oberseminar Marcus Völker
by Völker, Marcus 19 Jan '23

19 Jan '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 06. März 2023, 14:00 Uhr Ort: Raum 2202 (HBau, 2. Stock), Ahornstr. 55 Hybrid über Zoom: https://rwth.zoom.us/j/99712637671?pwd=U0UzV1JiL1hwWnlzNW5pUC9hVDdOUT09 Referent: Marcus Völker, M.Sc. RWTH Informatik 11 Embedded Software Thema: Policy Iteration for Value Set Analysis of PLC Programs Abstract: Ensuring the correct behaviour of computing systems is an important task to prevent danger to the users of such systems. To do this, many analysis techniques have been developed that can find bugs in software, or prove that it complies to some specification of correct behaviour. Among these techniques, a fundamental analysis is value set analysis (VSA), which can determine an approximation of the program variables' values at each point of the program. This is very important information, as many faulty behaviours can be traced back to variables taking unexpected values, such as division by zero, access to uninitialised memory or outside a buffer, or unreachable code. While classically, value set analysis is performed with the algorithm of Kleene iteration, another approach called policy iteration has been developed in recent years that provides an alternative with the potential of finding similar or better results than Kleene iteration in less time. Policy iteration works by using a heuristic to simplify the program in a certain way, finding the value sets of that program, and then checking whether the result is applicable to the original program. If yes, the results are used, otherwise different simplifications have to be checked, until a usable result is found. As policy iteration is a heuristic algorithm, it makes certain assumptions about program behaviour in order to achieve good results. It turns out, however, that these assumptions are not guaranteed if the program contains errors which cause it to behave differently than expected. As we use program analysis to find errors such as this, assuming an error-free program is not necessarily a good assumption. In this thesis, we show several ways to improve the original heuristic by focusing on program loops. First, we present a way to use a pre-analysis to determine some aspect of the loops' behaviours, and use this information in order to build a heuristic that leads to more accurate solutions than the standard heuristic in many cases, at the cost of additional running time necessary to perform this pre-analysis. Then, we show a way to reinterpret branches as loops if they occur in cyclical code, which is typical for programs in reactive systems, such as the systems used for factory automation. This allows us to use our loop heuristic on a wider variety of programs, even though the cost becomes even greater, and it is useful only in specific cases. Afterwards, we show how to remove the pre-analysis to gain back the lost time, while still retaining similarly good results to the expensive version introduced before. This has the additional benefit of allowing usage of the algorithms on branches as in the second approach, without incurring any additional costs. We motivate that this is the version of policy iteration that should be used in general with an extensive evaluation of generated programs. Finally, we show a way to analyse polynomial inequalities with value set analysis by reinterpreting them as conjunctions of simpler inequalities. This not only allows us to improve value set analysis results on programs that feature such inequalities, but also makes these programs accessible to policy iteration. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung Informatik-Oberseminar Sascha Müller
by Noll, Thomas 05 Jan '23

05 Jan '23

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * +********************************************************************** Zeit: Donnerstag, 12. Januar 2023, 10:00 Uhr Ort: 9222 (Gebäude E3, Informatikzentrum) Referent: Sascha Müller M.Sc. DLR Braunschweig Thema: Synthesizing FDIR Recovery Strategies for Space Systems Abstract: This talk proposes an inherently non-deterministic model for Dynamic Fault Trees (DFTs) to analyze Fault Detection Isolation and Recovery concepts with a particular focus on the needs of space systems. Deterministic recovery strategies are synthesized by transforming these non-deterministic DFTs into Markov automata. From the corresponding scheduler, optimized to maximize a given RAMS metric, an optimal recovery strategy can then be derived and represented by a model we call recovery automaton. We discuss dedicated techniques for reducing the state space of this recovery automaton and investigate lifting the approach to a partially observable setting. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung Informatik-Oberseminar: Beyond the rigid interfaces of
by Christos Psarras 22 Dec '22

22 Dec '22

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Monday, 23. January 2023, 15:30 Uhr Ort: Online (Zoom: https://umu.zoom.us/my/pauldj) Referent: Christos Psarras, M.Sc. International Research Training Group (IRTG-2379) Thema: Beyond the rigid interfaces of super-optimized building-block libraries. Our experiences in Chemometrics Abstract: The efficient computation of linear algebra expressions is a challenging task faced by many practitioners in scientific fields, such as engineering, image processing, and computational chemistry, to name a few. For most applications, mapping a target expression into a sequence of highly-optimized library routines (often referred to as "building-block" libraries, e.g., BLAS, LAPACK), is an approach that offers good computational performance as well as accuracy. However, in other applications, this approach inherently results in a vast under-utilization of the available computational resources, and thus reduced performance. In this talk, we emphasize on these, latter, applications, showcasing two occurrences that routinely arise in Chemometrics: the Canonical Polyadic Decomposition (CP) and Jackknife resampling of CP models. For the first occurrence, we describe the limitations of "mapping to building-blocks" when computing multiple, low-rank CP decompositions. After close collaboration with Chemometrics practitioners, we present a method (and algorithm), CP-CALS, which leverages information about their workflow, to overcome said limitations and achieve better performance. For the second occurrence, we describe the unique challenge of Jackknife resampling. We present a solution that addresses this challenge by making it possible to use CP-CALS to significantly increase performance, at the cost of slightly increasing the total amount of required computation. Through extensive experimentation with synthetic and real datasets on single-threaded and multi-threaded architectures, as well as on accelerators, we illustrate the improved efficiency and performance of our methods. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0