In this thread all special talks in our CogSys colloq will be announced.
Vortrag von: Corbinian Weithmann
Titel: Erklärbare Fehlerklassifikation in der bildbasierten industriellen Qualitätskontrolle
Zeit: 11.04.2025 - 15:00 - 16:00
Ort: WE5/05.013 und Teams (wegen Vertraulicher Information nur Audioübertragung)
Abstract folgt
Information to participate via Teams:
Microsoft Teams Benötigen Sie Hilfe?
Besprechungs-ID: 393 811 156 024
Kennung: CL6pv9QD
In reply to Felix Haase
Re: MA Verteidigung Corbinian Weithmann - Update mit Abstract
由Felix Haase發表於
Vortrag von: Corbinian Weithmann
Titel: Erklärbare Fehlerklassifikation in der bildbasierten industriellen Qualitätskontrolle
Zeit: 11.04.2025 - 15:00 - 16:00
Ort: WE5/05.013 und Teams (wegen Vertraulicher Information nur Audioübertragung)
Abstract:
In der Abteilung EA-343 der BMW Group wird eine Künstliche Intelligenz (KI) zur Fehlerdetektion bei der Entwicklung von Elektromotor-Komponenten eingeführt. Als Pilotkomponente wird der Stator verwendet, um die automatisierte Erkennung von Produktionsfehlern durch ein Convolutional Neural Network (CNN) zu testen.
Ziel ist die zuverlässige Identifikation von Fehlern, wie Blasen oder Risse, die manuell schwer zu erkennen sind. In einem zweistufigen Prozess entscheidet das CNN-Modell zunächst, ob ein Fehler vorliegt. Methoden der erklärbaren KI (xAI) werden verwendet, um die Vertrauenswürdigkeit und Qualität der Ergebnisse zu verbessern.
Erwartet wird, dass Ingenieure durch die automatisierte Fehlererkennung frühzeitig fundierte Entscheidungen treffen können, um Design und Prozesse zu optimieren.
Information to participate via Teams:
Microsoft Teams Benötigen Sie Hilfe?
Besprechungs-ID: 393 811 156 024
Kennung: CL6pv9QD
Presentation by: Adni Islami (University of Prishtina, Kosovo)
Title: Developing a Green Infrastructure Database for Prishtina: Insights from Research Experience in Würzburg
Time&Date: 14.05.2025 - 16:00 - 17:00
Location: WE5/05.013 and Teams
Abstract:
This presentation will focus on my current research work in Würzburg, which represents the first practical step within a broader project aimed at developing a green infrastructure database for the city of Prishtina. The overall goal of the project, funded by DBU, is to support the creation of a comprehensive tree cadastre to improve urban green space planning and management in Prishtina.
During my stay in Würzburg, I am working with local cadastral data on individual trees and exploring its structure in comparison with the publicly available dataset from the Bavarian open geodata portal. The aim is to understand how well-structured and accessible data systems function in practice, and how such models can inform and inspire the future development of tree data management in Prishtina.
Information to participate via Teams:
Microsoft Teams Benötigen Sie Hilfe?
Besprechungs-ID: 379 093 638 480
Kennung: jS3Mf7xf
There was an issue with the link to MS-Teams, this is the correct one:
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 379 093 638 480
Kennung: jS3Mf7xf
In reply to Felix Haase
MA Verteidigung Mai Anh Selina Vu - Integrating Large Language Models into Intelligent Tutoring Systems for Basic Arithmetic to Generate Text Exercises and Tailored Feedback
由Felix Haase發表於
Vortrag von: Mai Anh Selina Vu
Titel: Integrating Large Language Models into Intelligent Tutoring Systems for Basic Arithmetic to Generate Text Exercises and Tailored Feedback
Zeit: 11.06.2025 - 16:00 - 17:00
Ort: WE5/05.013 und Teams
Abstract:
This thesis investigates the integration of large language models (LLMs) into intelligent tutoring systems (ITS), with the goal of generating diverse and pedagogically meaningful subtraction-based text exercises for primary school students. Building on the existing ITS Subkraki – which previously relied on static, template-based exercises – the project introduces a dynamic content generation approach supported by a locally hosted LLM. Following a requirements-driven planning phase, the prototyping stage revealed key challenges, including long generation times and inconsistencies in the generated output, especially when using mid-sized models due to hardware constraints. To address these limitations, the system architecture was extended with a validation layer in form of an additional educator-facing frontend. This allows educators to generate, manually add, review, and correct exercises before they are made available to students, ensuring quality and reliability. The final implementation was evaluated against the previously defined requirements. Although the fully automated generation of consistently correct exercises that meet all set criteria is not yet feasible, the results demonstrate that LLMs can significantly enhance educational tools when combined with a human-in-the-loop workflow. Overall, the thesis provides a scalable and adaptable architecture that addresses the research question of how LLMs can be utilized to generate text exercises for ITS, while also managing challenges related to exercise difficulty, analogy selection, and prompt design within the constraints of the local deployment.
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 379 093 638 480
Kennung: jS3Mf7xf
In reply to Felix Haase
MA Verteidigung Philippe Crackau - Examining Image Classifier Decisions Through the Lense of Probabilistic Relational Explanations – Extending the CoReX Framework by Probabilistic Logic Programming
由Felix Haase發表於
Vortrag von: Philippe Crackau
Titel: Examining Image Classifier Decisions Through the Lense of Probabilistic Relational Explanations – Extending the CoReX Framework by Probabilistic Logic Programming
Zeit: 11.06.2025 - 13:00 - 14:00
Ort: WE5/05.013 und Teams
Abstract:
Explainability and human interpretability are relevant topics of research due to
the ubiquity of AI systems in various fields through the rise of Deep Neural Net-
works and the growing awareness of their shortcomings. The CoReX-framework
is one such system, which uses Prolog to learn human readable First-order logic
rules from features informing the NNs decision-making.
In this thesis, the current Prolog-explainer Aleph will be replaced with a prob-
abilistic inductive logic programming extension in Prolog by Riguzzi et al. in
order to gain an additional layer of information by weighting the certainty of
the classification.
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 383 659 694 423 3
Kennung: Mv68nY9K
In reply to Felix Haase
MA Verteidigung Sonja Ruschhaupt - An Intelligent Tutoring System for Primary-Level Tree Identification by Alignment and Socratic Dialogue
由Felix Haase發表於
Vortrag von: Sonja Ruschhaupt
Titel: An Intelligent Tutoring System for Primary-Level Tree Identification by Alignment and Socratic Dialogue
Zeit: 28.07.2025 - 09:00 - 10:00
Ort: WE5/05.013 und Teams
Abstract folgt
Information to participate via Teams:
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 392 396 124 216 2
Kennung: qd2Np2Tc
In reply to Felix Haase
Re: MA Verteidigung Sonja Ruschhaupt - An Intelligent Tutoring System for Primary-Level Tree Identification by Alignment and Socratic Dialogue
由Felix Haase發表於
Vortrag von: Sonja Ruschhaupt
Titel: An Intelligent Tutoring System for Primary-Level Tree Identification by Alignment and Socratic Dialogue
Zeit: 28.07.2025 - 09:00 - 10:00
Ort: WE5/05.013 und Teams
Abstract:
German primary teachers are overwhelmed with heterogeneity in classrooms leading to students not receiving the individual attention they require. Intelligent Tutoring Systems (ITSs) can relieve both teachers and students by providing feedback-rich independent learning opportunities to students, thus freeing up time for teachers for individual mentoring. This thesis introduces WaldSchwein, an ITS for teaching primary school students the identification and attribute description of trees. To do so, WaldSchwein conducts a socratic dialogue by dynamically adjusting tasks and responses to the student’s input. Contrasting images are provided to help the student align differences between classes upon incorrect classification. Additionally, a heuristic based on Laplace’s Rule of Succession (LRoS) for selecting the next classification problem presented to the student is developed as an alternative to Knowledge Tracing (KT) in the case of insufficient statistical data about tasks. The heuristic is tested using different prototype students. WaldSchwein’s usability was evaluated by experts in the fields of ITS and biology education. While the heuristic outperformed a random and ordered round-robin heuristic, its application to this particular teaching problem is likely unsuitable as breaks between sessions and outside influence muddy results. The usability study found that WaldSchwein was largely seen as a useful teaching device. In particular the use of contrasting images and abstraction found approval. The layout was said to need some improvement as in some task views information mapping from text to image was seen as too complicated. Overall, once the layout has been adjusted, WaldSchwein is ready for real student testing.
Information to participate via Teams:
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 392 396 124 216 2
Kennung: qd2Np2Tc
In reply to Felix Haase
MA Verteidigung Christoph Ubald Stößel - Ein multimodales interaktives Lehr-Lern-System zur Vermittlung von KI-Kompetenzen für KMUs
由Felix Haase發表於
Vortrag von: Christoph Ubald Stößel
Titel: Ein multimodales interaktives Lehr-Lern-System zur Vermittlung von KI-Kompetenzen für KMUs
Zeit: 31.07.2025 - 09:00 - 10:00
Ort: WE5/05.013 und Teams
Abstract:
Small and medium-sized enterprises (SMEs) represent a central business segment and play a crucial role for the German economy. Nevertheless, the increasingly important topic of artificial intelligence (AI) remains underrepresented by most SMEs, primarily due to a lack of necessary skills. This can be attributed to various inhibiting factors, one of which is the lack of access to information materials addressing the needs of the target group. As part of this study, a multimodal interactive tutor system was developed, to inform SME employees about AI and its practical applications, while also aiming to build relevant competencies. The content is tailored to the target group based on learning theories from adult education and multimedia learning. Furthermore, special emphasis is placed on ensuring a high level of usability. The development followed a design-based research approach, encompassing the phases of requirements analysis, implementation and evaluation. The system’s usability was evaluated using the System Usability Scale (SUS). Additional insights into content design were gained through qualitative supplementary questions. The results indicate that the system positively influences user perception, thanks to its wide range of interactive features and clear visual design. Moreover, the evaluation results offer valuable insights for potential improvements in a subsequent iteration of the system.
Information to participate via Teams:
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 334 724 880 415 0
Kennung: qv97Uf75
In reply to Felix Haase
MA Zwischenpräsentation Amir Ebrahimi Moghaddam - Revising CNN Models via Augmented Near Hits and Near Misses in Binary Image Classification
由Felix Haase發表於
Vortrag von: Amir Ebrahimi Moghaddam
Titel: Revising CNN Models via Augmented Near Hits and Near Misses in Binary Image Classification
Zeit: 30.07.2025 - 16:00 - 17:00
Ort: WE5/05.013 und Teams
Abstract:
In this thesis, we propose a method for revising a baseline CNN model by augmenting images selected through Near Hits and Near Misses (NHMs) found in its misclassified predictions. Unlike traditional random augmentation, our approach aims to target the decision boundary and support more meaningful model improvements. We experiment with different strategies for selecting misclassified images: randomly or based on low-confidence predictions close to the decision boundary. These misclassified samples are then used to find NHMs( visually or semantically similar examples) which are augmented and added back into training. We compare the effects of NHM-based augmentation with random selection, measuring improvements in accuracy and generalization. Key challenges include ensuring diversity in augmentation, choosing the right model capacity, and determining how misclassified sample selection affects NHM quality. Our results aim to reveal whether model uncertainty and semantic similarity can guide more effective and interpretable model revisions.
Information to participate via Teams:
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 379 093 638 480
Kennung: jS3Mf7xf
In reply to Felix Haase
MA Verteidigung Kemal Tanirak - "Leben in Deutschland" als Serious Game
由Felix Haase發表於
Vortrag von: Kemal Tanirak
Titel: "Leben in Deutschland" als Serious Game: Vermittlung von deklarativem Wissen durch ein dialogbasiertes System auf Basis eines LLMs (MA CitH)
Zeit: 09.09.2025 - 14:00 - 15:00
Ort: WE5/05.005 und Teams
Abstract folgt
Information to participate via Teams:
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 344 663 060 513 1
Kennung: Yo7Qa7MW
In reply to Felix Haase
MA Verteidigung Max Brombach - Automated Test Case Generation for Bug Detection in Python Programs Using Large Language Models
由Felix Haase發表於
Vortrag von: Max Brombach
Titel: Automated Test Case Generation for Bug Detection in Python Programs Using Large Language Models: A Comparative Study of End-to-End and Systematic Test Case Generation
Zeit: 27.08.2025 - 12:00 - 13:00
Ort: WE5/05.005 und Teams
Abstract folgt
Information to participate via Teams:
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 340 579 583 586 4
Kennung: 8aQ3Fp9V
In reply to Felix Haase
MA Verteidigung Eva Jansohn - An Intelligent Tutoring System for Solving Multiplication Word Problems with Error-specific Feedback by Analogical Worked-out Examples
由Felix Haase發表於
Vortrag von: Eva Jansohn
Titel: An Intelligent Tutoring System for Solving Multiplication Word Problems with Error-specific Feedback by Analogical Worked-out Examples
Zeit: 24.09.2025 - 16:00 - 17:00
Ort: WE5/05.005 und Teams
Abstract folgt
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 379 093 638 480
Kennung: jS3Mf7xf
In reply to Felix Haase
Probevorträge zur KI Konferenz 2025 zu kontrastiven Erklärungen
由Felix Haase發表於
Vortrag: Yuliia Kaidashova
Titel: Toward Short and Robust Contrastive Explanations for Image Classification by Leveraging Instance Similarity and Concept Relevance
Vortrag: Luisa Schneider
Titel: Re-Evaluating the Robustness and Interpretability of the Contrastive Explanation Method for Image Classification
Zeit: 10.09.2025 - 16:00 - 17:00
Ort: WE5/05.005 und Teams
________________________________________________________________________________
Microsoft Teams Benötigen Sie Hilfe?
Jetzt an der Besprechung teilnehmen
Besprechungs-ID: 379 093 638 480
Kennung: jS3Mf7xf
In reply to Felix Haase
Re: Probevorträge zur KI Konferenz 2025 zu kontrastiven Erklärungen
由Felix Haase發表於