+**********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+**********************************************************************
Zeit:
Dienstag, 30. Januar 2024, 14.00 Uhr
Ort:
9222, E3, Ahornstr. 55 und hybrid via Zoom (https://rwth.zoom-x.de/j/64937773189?pwd=eGttNUMzSElnQUVkc3FrYzBqK2F4UT09)
Referent:
Lubna Ali M.Sc. RWTH
Lehr- und Forschungsgebiet Informatik 9 (Lerntechnologien)
Thema:
convOERter: A Technical Assistance Tool to Support Semi-Automatic Conversion of Images in Educational Materials as OER
Abstract:
Open Educational Resources (OER) are seen as an important element in the process of digitizing higher education teaching and as essential building blocks for openness in education. They can be defined as teaching, learning, and research materials that have been made openly available, shareable, and modifiable. OER include different types of resources such as full courses, textbooks, videos, presentations, tests, and images, which are usually published under the open Creative Commons licences. OER can play an important role in improving education by facilitating access to high quality digital educational materials. Accordingly, there is a steady increase among higher education institutions to participate in the so-called "open movement" in general and in utilizing OER in particular. Nevertheless, there are many challenges that still face the deployment of OER in the educational context. One of the main challenges is the production of new OER materials and converting already existing materials into OER, which could be viable by qualifying educators through training courses and/or supporting them with specific tools.
There are many platforms and tools that support the creation of new OER content. However, to our knowledge, there are no tools that perform fully- or semi-automatic conversion of already existing educational materials. This identified gap was the basis for the design and implementation of the OER conversion tool (convOERter). The tool supports the user by semi-automatically converting educational materials containing images into OER-compliant materials. The main functionality of the tool is based on reading a file, extracting all images as well as all possible metadata, and substituting the extracted images with OER elements in a semi-automated way. The retrieved OER images are referenced and licenced properly according to the known TASLL rule. Finally, the entire file is automatically licenced under Creative Commons excluding specific elements from the entire licence such as logos. In order to evaluate the effectiveness of the tool in promoting the use of OER, a comprehensive user study was conducted with educators and OER enthusiastic at different universities. The study was accomplished by offering a series of OER evaluation workshops to compare the conversion efficiency of the tool with manual conversion. The results show that using the conversion tool improves the conversion process in terms of speed, license quality, and total efficiency. These results highlight that the tool can be a valuable addition to the community, especially for users less experienced with OER. As a future work, it is intended to further develop the tool and improve its functionality. Additionally, a long-term study can be conducted to assess the impact of the tool in facilitating and enhancing the production of OER on a larger scale.
Es laden ein: die Dozentinnen und Dozenten der Informatik
+***********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+***********************************************************************
Zeit: Montag, 4. November 2024, 14:30 Uhr
Ort: Raum 9222 (Informatikzentrum E3, Ahornstraße 55)
Referent: Jan Martin Tönshoff, M.Sc.
Lehrstuhl für Logik und Theorie diskreter Systeme (Informatik 7)
Thema: Deep Learning on Graphs: Developing and Understanding Neural Architectures for Structured Data
Abstract:
Graph Neural Networks (GNNs) have recently emerged as a powerful class of deep
learning architectures that can directly learn from graph-structured data. Such
data naturally occurs in a wide range of structurally rich learning domains, such
as computational chemistry or social network analysis. This thesis aims to study
both practical and theoretical aspects of GNNs, with a focus on understanding
their capabilities for solving hard learning tasks, exploring new directions for GNN
architecture design, and understanding the relative strengths and weaknesses of
commonly used models. In this context, we present our main research contributions:
Firstly, we propose a GNN-based approach for learning heuristics for Constraint
Satisfaction Problems (CSPs) through a general graph representation and neural
architecture. Using reinforcement learning, we show that this approach can learn
powerful search algorithms, achieving superior performance compared to prior
neural methods and being competitive with classical solvers on a range of complex
combinatorial problems.
Secondly, we examine a novel GNN architecture based on 1D convolutions on
sequential features constructed from random walks. We analyze the expressivity
of this approach and prove it to be incomparable to the Weisfeiler-Leman hierarchy
and many common GNN architectures. The model is further shown to achieve
strong empirical performance across various tasks, suggesting a new direction for
GNN architecture design beyond traditional message passing.
Thirdly, we compare popular GNNs in the context of capturing long-range
dependencies in graphs. Through an empirical re-evaluation of commonly used
benchmark datasets, we show that standard GNNs based on message passing
perform better than previously reported. A theoretical comparison between Graph
Transformers and GNNs with Virtual Nodes further shows that neither architecture
subsumes the other in terms of uniform expressivity, highlighting the need to
consider a wide range of architectures for specific learning tasks.
Es laden ein: die Dozentinnen und Dozenten der Informatik
+**********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+**********************************************************************
Zeit: Mittwoch, 30. Oktober 2024, 14:00 Uhr
Ort: Im Süsterfeld 9, 52072 Aachen, Raum 273
Referent: Sebastian Hueber M.Sc.
Lehrstuhl Informatik 10
Thema: Using Facial Tracking for Expressive Mobile Device Interactions
Abstract:
For four decades, user interfaces have been mainly designed for pointing input, either with a mouse or a touchscreen. Pointing input abstracts from the human to a location on the screen. Especially visual cues that make human-to-human communication effective – like eye contact or head nodding – are ignored in this abstraction. Mobile devices, in particular, suffer from the limits of pointing input, as users cannot comfortably reach everything on the screen when using the device one-handedly. We present implicit and explicit usages of facial tracking to make mobile interactions more expressive and ergonomic.
We show the advantages of eye tracking using three interaction techniques. First, our Attentive Notifications remove occlusion issues and accidental activations in mobile interfaces. They determine a suitable screen edge for displaying notifications by blocking the area around the user’s gaze at the moment of notification delivery. Second, we show that eye tracking can enhance the perception of content in augmented reality with our User-Aware Rendering. This technique provides enhanced depth perception with good performance in scene exploration. Third, interfaces can exploit that gaze input can reach anything nearby. Our GazeConduits concept fosters collaboration in ad-hoc multi-device environments. This enables users to interact with devices or meeting collaborators by looking at them.
However, eye tracking often comes with accuracy issues, especially when people are moving, and suffers from the Midas touch problem. To overcome these challenges, two of our interaction techniques use head tracking instead. We present a Head + Touch controlled cursor that increases the thumb’s reach during one-handed smartphone use. This significantly reduces the overhead of touch-based reachability techniques to under 100 ms. With our Headbang technique, menu selections are also faster than with touch input.
Es laden ein: die Dozentinnen und Dozenten der Informatik
+**********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+**********************************************************************
Zeit: Freitag, 11. Oktober 2024, 15:30 Uhr
Ort: Raum 5053.2 (großer B-IT-Hörsaal), Informatikzentrum, E2, Ahornstr. 55
Referentin: Bahare Salmani Barzoki, M.Sc.
Lehrstuhl für Softwaremodellierung und Verifikation (Informatik 2)
Thema: Probabilistic Model Checking and Parameter Tuning for Bayesian
Networks
Abstract:
Probabilistic reasoning is a key to handling uncertainties and making
decisions based on partial observations. Bayesian networks (BNs) are
popular probabilisticgraphical models in probabilistic decision-making
and AI and have a wide range of applications, including machine
learning, medicine, gene regulatory networks, and robotics. They combine
the notions from probability theory and graph theory and enablea
succinct representation of joint probability distributions.A Bayesian
network is composed of a directed acyclic graph over a set of random
variables and a set of conditional probability tables ---CPTs for short.
The primary task in Bayesian networks is probabilistic inference, that
is, to compute a conditional probability of a joint valuation for a
subset of random variables given an observation.
Probabilistic model checking is a field in computer science that
isfocusedon analyzing stochastic systems with respect toa set of
formally defined properties. The stochastic systems are typically Markov
models, and the properties of interest are probabilistic extensions of
LTL of CTL. The properties are mostly reducedto a pivotal task,
computing reachability probabilities: whatis the probability of reaching
a set of target states? Recent advancements in the field consider the
parametric extensions of the Markov models, where a subset of the
probabilities in the model are unknown and target various synthesis
problems including/feasibility checking/: is there a satisfying
instantiation that satisfies the given constraint?,/region
verification/: are all instantiations in a /region/ satisfy the given
constraint, and /parameter space partitioning/: split the entire
n-dimensional parameter space to sets of satisfying, rejecting, and
unknown subregions w.r.t. the given constraint.
In this dissertation, we present a new approach based on /probabilistic
model checking/ for inference and parameter tuning in Bayesian networks.
The dissertation is categorizedinto two main settings: the
/non-parametric/ and the /parametric/. In the non-parametric setting,we
focus on the classical Bayesian networks, where the model is fully
specified.We present mappings from Bayesian networks to discrete-time
Markov chains and mathematically reduce performing conditional inference
in the BN to computing reachability probabilities in Markov chains.
Thisenables the use of state-of-the-art algorithms for computing
reachability probabilities and the optimization techniques thereof,
e.g., bisimulation minimization for exact inference in BNs. We exploit
the explicit and symbolic methods from probabilistic model checking and
empirically evaluate our framework for Bayesian inference against the
state-of-the-art /weighted model counting. /In the parametric setting,we
define /parametric Bayesian networks (pBNs)/, where a subset of CPT
entries are unknown polynomials rather than concrete probabilities.We
build upon the synthesis techniques for parametric Markov chains and
address problems /sensitivity analysis/, /ratio/difference parameter
tuning/ from the BN literature, and /parameter space partitioning/from
pMC literature. Finally, we focus on the problem
/minimal-distanceparameter tunin/g, where the objective is to find the
instantiationu for the unknown parameters that satisfy the constraint of
interest while minimally deviating from the original instantiation u_0
in an original reference model. The motivation is to minimally disturb
the statistical information in the original model.
Our detailed experimental evaluations indicate that our parameter
synthesis techniques can treat parameter synthesis for Bayesian networks
(with hundreds of unknown parameters) that go beyond the capabilities of
the existing techniques. We lift the severe restrictions in the
literature on the number of unknown parameters, the global dependencies
between the parameters, and the form of obtained solutions.
------------------------------------------------------------------------
Es laden ein: die Dozentinnen und Dozenten der Informatik
_______________________________________________
informatik-sekretariate Mailingliste --
informatik-sekretariate(a)lists.rwth-aachen.de
Um den Newsletter abzubestellen, senden Sie bitte eine E-Mail an
informatik-sekretariate-leave(a)lists.rwth-aachen.de