Offene Abschlussarbeiten

Description: The growing adoption of artificial intelligence (AI) has created a need for models that are not only accurate but also interpretable. Explainable AI (XAI) aims to provide insights into how AI systems make decisions, which is critical for enhancing transparency, accountability, and trustworthiness. However, measuring and evaluating the explainability of AI models remains a challenge. This thesis will focus on exploring the testing and evaluation of explainability in AI. The goal is to develop a systematic approach to assess the degree of explainability of different AI models and identify best practices for designing explainable AI systems. The thesis will involve conducting a comprehensive literature review, designing and conducting experiments to evaluate the explainability of AI models, and proposing metrics to quantify the interpretability of models.

Description: Image segmentation and quality inspection are crucial tasks in various domains, such as manufacturing, healthcare, and autonomous vehicles. Deep learning-based approaches have shown remarkable performance in these tasks; however, they often lack interpretability, making it challenging to understand how the models make decisions. Explainable AI (XAI) can help address this issue by providing insights into the decision-making process of AI models. This thesis will focus on developing an XAI approach for image segmentation and quality inspection tasks. The goal is to design and implement an XAI system that can accurately segment images and inspect their quality while providing interpretability into the model’s decision-making process. The thesis will involve conducting a comprehensive literature review, designing and implementing a deep learning-based image segmentation and quality inspection model, developing an XAI approach to explain the model’s decision-making process, and evaluating the effectiveness of the XAI system in a given use case based on the TTPLA dataset. The thesis will also involve exploring the potential of integrating different XAI techniques, such as attention mechanisms and visualization, to enhance the interpretability of the model. This research will contribute to the advancement of XAI and help organizations develop more interpretable and trustworthy image segmentation and quality inspection systems.

Description: As AI systems are increasingly integrated into various domains, there is a growing need for interpretable AI models that can explain their decision-making processes. However, designing and implementing explainable AI (XAI) systems can be a complex and time-consuming process. Therefore, the development of a generalizable XAI platform can significantly accelerate the deployment of interpretable AI models across different domains. This thesis will focus on developing an approach for a generalizable XAI platform. The goal is to design and/or implement an XAI platform that can be used across multiple domains and scenarios, enabling developers to integrate explainable AI models into their systems more easily. The thesis will involve conducting a comprehensive literature review, identifying common challenges and requirements for XAI platforms, designing and/or implementing a prototype platform, and evaluating its usability and effectiveness in different use cases. The thesis will also involve exploring the potential of integrating different XAI techniques, such as rule-based models, decision trees, and visualizations, into the platform. This research will contribute to the advancement of XAI and help organizations develop more interpretable and trustworthy AI systems.

Context: Large Language Models like GPT-3 or BERT have become increasingly popular within the last years. Due to the strong improvements in natural language understanding and generation they have a huge potential to be used in real-world applications within  a broad spectrum of application fields. However, it was shown that they can adapt unethical behavior, e.g. by using inappropriate language or stating wrong information.

Task: Develop a framework to systematically assess the trustworthiness of Large language models.

Please apply with brief motivation letter and resume.

Context: Trustworthy AI refers to the development and deployment of artificial intelligence systems that are reliable, ethical, transparent, and secure. It aims to ensure that AI systems are developed and used in a way that respects human rights and values, is free from bias and discrimination, and can be trusted by individuals and society as a whole.

Task: Development of a Framework to systematically assess the trustworthiness of a ML learning model based on text or tabular data.

Please apply with brief motivation letter and resume.

Context: Trustworthy AI refers to the development and deployment of artificial intelligence systems that are reliable, ethical, transparent, and secure. It aims to ensure that AI systems are developed and used in a way that respects human rights and values, is free from bias and discrimination, and can be trusted by individuals and society as a whole.

Task: Development of a concept for requirements engineering for Trustworthy AI.

Please apply with brief motivation letter and resume.

Context: Synthetic data is computer-generated data that can be used to supplement or replace real-world data in applications like machine learning and data analysis. It can help overcome limitations in available data and protect privacy.

Task: Systematic Literature Review on Techniques to generate synthetic Data

Please apply with brief motivation letter and resume.

Anwendungen können auf vielfältige Weise entwickelt und betrieben werden. Von Monolithen On-Premises bis zu Serverless und Cloud-native Microservices ist vieles möglich, wenn auch nicht immer alles sinnvoll. Jedes Pattern hat seine eigenen Herausforderungen bezüglich:

• Kosten
• Komplexität
• Performance
• Skalierbarkeit
• Security
• Deployment
• Testing und vielen weiteren

Das Ziel der Abschlussarbeit ist die Analyse dieser Faktoren am konkreten Anwendungsfall GovRadar. Dieses im Frühjahr 2020 gegründete Tech-Startup treibt die Digitalisierung Deutschlands voran, indem es modernste Technologien einsetzt, um etablierte Prozesse in Behörden zu automatisieren oder komplett abzulösen. Aktuell wird u.a. eine innovative, datengetriebene Beschaffungsplattform für den öffentlichen Sektor entwickelt.

Wir suchen stets engagierte Kolleginnen und Kollegen mit agilem Mindset, die dafür brennen, herausragende Softwarelösungen zu entwickeln und die mit uns den XITASO Weg mitgestalten möchten. Bei XITASO kannst Du Dein im Studium erworbenes Wissen direkt praktisch anwenden. Die Einarbeitung in das Thema verschafft Dir sowohl Grundlagen- als auch vertiefte Kenntnisse aus der Praxis. Zur Ausschreibung: https://xitaso.com/karriere/jobs/abschlussarbeit-cloud-architekturen-und-design-patterns/

Bewerbungen bitte mit kurzem Anschreiben (z.B. Erfahrungen im Bereich Softwareentwicklung) und Lebenslauf.

Methoden des Maschinellen Lernens (ML) werden heutzutage in zahlreichen Anwendungsgebieten eingesetzt. In Anbetracht der realen (gesellschaftlichen) Auswirkungen besteht ein hoher Anspruch an die Vertrauenswürdigkeit von ML-basierten Systemen. Das Forschungsgebiet „Vertrauenswürdige KI“ (engl. Trustworthy AI) untersucht, wie ein sicherer, transparenter und verantwortungsvoller Einsatz von KI bzw. ML gestaltet werden kann. Dabei sind verschiedene Anforderungen zu erfüllen. Ein wichtiger Aspekt ist, dass Vorhersagen von ML-Systemen fair sind, d.h. keine bestimmten Personen(gruppen) diskriminiert werden. Zum anderen sollen ML-Systeme robust gegenüber Abweichungen realer Daten von Trainingsdaten sein (engl. „Distribution Shift“, z.B. verschiedene Lichtverhältnisse in Bildern). Mögliche Themen:

• Definition und Evaluation von Robustheit: Distribution Shift, Reliability, Uncertainty
• Umfragen oder Experteninterviews zum Stand in KI entwickelnden Unternehmen bzgl. Umgang mit Bias, Distribution Shift in ML-Anwendungen

Bewerbungen bitte mit Themeninteresse/ kurzer Motivation und Lebenslauf.

Maschinelles Lernen ist ein Thema von großem Forschungsinteresse.

Für aktuelle Trends siehe z.B. Google AI Blog.

Bitte nur mit einem eigenen Thema bzw. konkreter Forschungsfrage bewerben.

Beispiel: CO2 Fußabdruck von ML und energiesparende Maßnahmen

Computer Vision is a subfield of Machine Learning (ML) that typically uses Convolutional Neural Networks (CNNs) to process Images. Instead of using CNNs, Google recently suggested new archtectures (see e.g., Transformer and Mixer) to deal with spatial data. While the top-line generalization accuracy has been comparatevly to CNNs, the fairness (measured as the parity of accuracy) among subgropus remains an open question.

Using the publicly available CelebA dataset, the goal of this thesis is to compare ResNet18 (representative for CNNs), ViT, and MLPMixer in terms of their accuracy within ethnical subgroupus (i.e., Gender and Age). See Sagawa et al. (2020) for the methodological setup (also code snippets available upon request) and Hooker et al. (2020) for a rigorous assessment of algorithmic bias.

NVIDIA GPU or Google Colab Prenium recommended.

• Grundkenntnisse in Python oder anderen Programmiersprachen
• Grundkenntnisse oder Interesse an ML-Modellen, insbesondere im Bereich Natural Language Processing

Graph Neural Networks (GNNs) have gained interest in the past years, as they provide a suitable neural networks based technique to process graph and network data. Typical graph data are social networks, used in platforms like Twitter and Facebook.Recently, there have been efforts to combine GNNs with Reinforcement Learning with promising results. Exemplary use cases are fraud detection and molecular property prediction (E.g. for medicinal drug discovery).The aim of this work is to provide a comprehensive systematic literature analysis for GNNs in combination with Reinforcement Learning.

Neural Recommendation System for Academic Contents – A Graph Neural Network Based Approach at FAU (VERGEBEN)

Recommendation Systems have been successfuly used in many areas, like marketing and E-commerce. A reason is that this technology tackles the problem of information overload, which often occurs in decision processes.

For this thesis, the aim is to implement a prototype for a recommendation algorithm for the academic content at the FAU with the usage of Graph Neural Networks. These type of deep learning technique are used in many state of the art recommendation approaches (e.g. recommendations at pinterest: shorturl.at/hFNP1).  Data (Already preprocessed) and a first graph structure will be provided.

Prerequisites:

• Experience in programming languages like python and machine learning frameworks
• Interess or experience in graph based approaches