Forschungsinteressen
Meine Forschung verbindet datengestützte Modellierung mit physikalischen Prinzipien, um robuste und konsistente Systemmodelle zu entwickeln.
Methodische Schwerpunkte
- Physik-informierte Ansätze: Physics-Informed Neural Networks (PINNs) mit physikalischen Randbedingungen; Diffusion Models zur physikalisch plausiblen Generierung stochastischer Prozesse
- Generative Modelle: GANs sowie Autoencoder (VAEs/AE) zur Analyse latenter Systemzustände und Modellierung komplexer Datenverteilungen
- AI Agents & Autonome Systeme: Entwicklung intelligenter Agenten für dynamische Umgebungen; Steuerung via Reinforcement Learning auf Basis von Bellman- und Hamilton-Jacobi-Bellman-Prinzipien
- Interpretierbarkeit: XAI-Methoden für komplexe Architekturen (Transformer, EvoFormer) wie auch in Computer Vision; Random Forests als Brücke zwischen Vorhersagekraft und Transparenz
- NLP & Statistische Analyse: Syntaxanalyse und Informationsextraktion für Muster- und Anomalieerkennung in unstrukturierten Textdaten
- Data Science & Datenanalyse: Explorative Datenanalyse, Feature Engineering und statistische Modellierung zur Extraktion relevanter Muster aus komplexen, hochdimensionalen Datensätzen
Entwicklung hybrider Modelle, die physikalische Fundierung mit der Flexibilität neuronaler Netze vereinen – wissenschaftlich präzise, inhaltlich konsistent.
Themen für Abschlussarbeiten
Abschlussarbeiten gerne auf Anfrage - Breite/Induviduelle Themenauswahl im Bereich Data Science und Machine Learning (KI)
Betreute Arbeiten
| Abschluss | Grad | Titel |
|---|---|---|
| 2026/02 | Bachelor | Gated Recurrent Unit Cells for Resource Constrained Devices: Concept, Implementation and Evaluation |
| 2026/02 | Bachelor | Unsicherheit und Vertrauensmaße: Eine Gegenüberstellung neuronaler und probabilistischer Verfahren |
| 2026/02 | Bachelor | Neural Architecture Search – Automatic CNN Design for Medical Images |
| 2026/02 | Bachelor | Analysis and Comparison of Ensemble Learning Methods for Contract Data Quality Assessment in Automated Order Fulfillment Process |
| 2026/02 | Bachelor | Attention-Mechanismus für multivariate Zeitreihen – Interpretierbare Sensorik-Analysen |
| 2026/02 | Bachelor | Enhancing cyberattack detection through cross-dataset analysis of unsupervised anomaly detection models |
| 2026/01 | Bachelor | Vergleich von Dimensionsreduktionsverfahren |
| 2025/12 | Master | Edge AI-Based Face Detection: Deployment and Optimization of Deep Learning Models on the ESP32-S3 Sense |
| 2025/12 | Bachelor | Simplification of Scientific Papers using RAGs in LLMs |
| 2025/10 | Bachelor | Enhancing Retail Decision Making - Interpretable Sales Forecasting using SHAP and “What if” Analysis |
| 2025/10 | Bachelor | Classification of Simulated Stenosis Degrees Using Remote Photoplethysmography and Machine Learning |
| 2025/10 | Bachelor | Architektur trifft Algorithmus: Baustilklassifikation sakraler Architektur mittels Entscheidungsbaum – und DBSCAN-Algorithmus |
| 2025/10 | Bachelor | Fraud Analysis on Imbalanced Credit Card Data Using Resampling, Classification and Anomaly Detection |
| 2025/09 | Master | Suchmaschinenoptimierung unter Systemrestriktionen: Eine Fallstudie am Beispiel einer Stand-up-Comedy Veranstaltungswebsite |
| 2025/08 | Bachelor | Secure CAPTCHA Generation via Stylized Symbol Synthesis with Generative AI |
| 2025/08 | Bachelor | Reassessing Classical Chinese Translation: A BERT-Based Approach with Custom Semantic Similarity Evaluation |
| 2025/08 | Bachelor | A systematic approach: Proposition of a hybrid model of CNN and ViT for Alzheimer’s classification and explainability |
| 2025/08 | Bachelor | Datenbasierte Prognose von Verkaufsprovisionen im Gebrauchtwagenhandel |
| 2025/07 | Bachelor | Data-Driven Anomaly Detection in Access Control for AI-Based Fraud Prevention |
| 2025/04 | Bachelor | Stock Price Prediction: Gaussian Processes versus Wassterstein-GANs |
| 2025/04 | Bachelor | LLM-based Automatization of Job-Position-Matching |
| 2025/04 | Bachelor | Vision-basierte Steuerung in Echtzeit-Kampfspielen mittels Deep Q-Networks |
| 2025/04 | Bachelor | Dynamic Obstacle Avoidance: Comparing Reinforcement Learning and Imitation Learning in Unity |
| 2025/03 | Bachelor | Von Affen mit der Schreibmaschine zum intelligenten Agenten: Wortgenerierung mit Q-Learning |
| 2025/02 | Bachelor | Explainable Transformer Architecture for Word-Level Sign Language Classification |
| 2025/01 | Bachelor | Learning chess using Double Deep Q-Networks (DDQN) |
| 2025/01 | Bachelor | Optimal Placement of Electrical Vehicle Chargers based on real world historic data: Genetic algorithm approach |
| 2024/12 | Bachelor | Active Semi-Supervised Learning for Improved CVD Classification |
| 2024/12 | Bachelor | Formel1 fahren Vergleich von PPO und SAC |
| 2024/12 | Bachelor | Predicting player churn in League of Legends using tree-based machine learning methods |
| 2024/11 | Bachelor | Natural Language Processing for Assigning Requirements to Verification Domains |
| 2024/09 | Bachelor | Using transfer learning (Fine-tuning vs Feature Extraction) on DestilBERT for string-matching |
| 2024/07 | Bachelor | Evaluation of the vulnerability of the deepfake detector to Grad-CAM-based adversarial attacks |
| 2024/07 | Bachelor | Automated Detection of Obstacles in the Cable-Channel of a Shearer Loader |
| 2024/06 | Bachelor | Feature Importance Random Forest |
| 2024/06 | Bachelor | Exploring User Interaction Metrics for Enhanced Web Content Evaluation |
| 2024/05 | Bachelor | NLP-basierte Stimmungsanalyse: Methoden und Anwendungen |
| 2024/05 | Bachelor | Street Lamp Area Detection through Classical Vision and Convolutional Neural Network |
| 2024/01 | Bachelor | Object Detection and Model Attention with YOLO (D-RISE) |
| 2023/12 | Bachelor | Combination of Grid Search and Random Search for more efficient hyperparameter optimization of neural networks. |
| 2023/12 | Bachelor | Transformer-based alignment and parsing of entities with semi-structured documents |
- (2026/02, BA) Gated Recurrent Unit Cells for Resource Constrained Devices: Concept, Implementation and Evaluation
- (2026/02, BA) Unsicherheit und Vertrauensmaße: Eine Gegenüberstellung neuronaler und probabilistischer Verfahren
- (2026/02, BA) Neural Architecture Search – Automatic CNN Design for Medical Images
- (2026/02, BA) Analysis and Comparison of Ensemble Learning Methods for Contract Data Quality Assessment in Automated Order Fulfillment Process
- (2026/02, BA) Attention-Mechanismus für multivariate Zeitreihen – Interpretierbare Sensorik-Analysen
- (2026/02, BA) Enhancing cyberattack detection through cross-dataset analysis of unsupervised anomaly detection models
- (2026/01, BA) Vergleich von Dimensionsreduktionsverfahren
- (2025/12, MA) Edge AI-Based Face Detection: Deployment and Optimization of Deep Learning Models on the ESP32-S3 Sense
- (2025/12, BA) Simplification of Scientific Papers using RAGs in LLMs
- (2025/10, BA) Enhancing Retail Decision Making - Interpretable Sales Forecasting using SHAP and “What if” Analysis
- (2025/10, BA) Classification of Simulated Stenosis Degrees Using Remote Photoplethysmography and Machine Learning
- (2025/10, BA) Architektur trifft Algorithmus: Baustilklassifikation sakraler Architektur mittels Entscheidungsbaum – und DBSCAN-Algorithmus
- (2025/10, BA) Fraud Analysis on Imbalanced Credit Card Data Using Resampling, Classification and Anomaly Detection
- (2025/09, MA) Suchmaschinenoptimierung unter Systemrestriktionen: Eine Fallstudie am Beispiel einer Stand-up-Comedy Veranstaltungswebsite
- (2025/08, BA) Secure CAPTCHA Generation via Stylized Symbol Synthesis with Generative AI
- (2025/08, BA) Reassessing Classical Chinese Translation: A BERT-Based Approach with Custom Semantic Similarity Evaluation
- (2025/08, BA) A systematic approach: Proposition of a hybrid model of CNN and ViT for Alzheimer’s classification and explainability
- (2025/08, BA) Datenbasierte Prognose von Verkaufsprovisionen im Gebrauchtwagenhandel
- (2025/07, BA) Data-Driven Anomaly Detection in Access Control for AI-Based Fraud Prevention
- (2025/04, BA) Stock Price Prediction: Gaussian Processes versus Wassterstein-GANs
- (2025/04, BA) LLM-based Automatization of Job-Position-Matching
- (2025/04, BA) Vision-basierte Steuerung in Echtzeit-Kampfspielen mittels Deep Q-Networks
- (2025/04, BA) Dynamic Obstacle Avoidance: Comparing Reinforcement Learning and Imitation Learning in Unity
- (2025/03, BA) Von Affen mit der Schreibmaschine zum intelligenten Agenten: Wortgenerierung mit Q-Learning
- (2025/02, BA) Explainable Transformer Architecture for Word-Level Sign Language Classification
- (2025/01, BA) Learning chess using Double Deep Q-Networks (DDQN)
- (2025/01, BA) Optimal Placement of Electrical Vehicle Chargers based on real world historic data: Genetic algorithm approach
- (2024/12, BA) Active Semi-Supervised Learning for Improved CVD Classification
- (2024/12, BA) Formel1 fahren Vergleich von PPO und SAC
- (2024/12, BA) Predicting player churn in League of Legends using tree-based machine learning methods
- (2024/11, BA) Natural Language Processing for Assigning Requirements to Verification Domains
- (2024/09, BA) Using transfer learning (Fine-tuning vs Feature Extraction) on DestilBERT for string-matching
- (2024/07, BA) Evaluation of the vulnerability of the deepfake detector to Grad-CAM-based adversarial attacks
- (2024/07, BA) Automated Detection of Obstacles in the Cable-Channel of a Shearer Loader
- (2024/06, BA) Feature Importance Random Forest
- (2024/06, BA) Exploring User Interaction Metrics for Enhanced Web Content Evaluation
- (2024/05, BA) NLP-basierte Stimmungsanalyse: Methoden und Anwendungen
- (2024/05, BA) Street Lamp Area Detection through Classical Vision and Convolutional Neural Network
- (2024/01, BA) Object Detection and Model Attention with YOLO (D-RISE)
- (2023/12, BA) Combination of Grid Search and Random Search for more efficient hyperparameter optimization of neural networks.
- (2023/12, BA) Transformer-based alignment and parsing of entities with semi-structured documents
Lehre
WiSe 25/26
SoSe 2025
SoSe 2023
Gérald Kämmerer
Wissenschaftlicher Mitarbeiter
Kontakt
- Telefon: +49 203 37 93482
- E-Mail: gerald.kaemmerer[@uni-due.de]
Postadresse
- Gebäude BC, Raum 412
- Bismarckstr. 90 47057 Duisburg, Deutschland
Sprechzeiten
Nach Vereinbarung