Götz-Henrik Wiegand

Götz-Henrik Wiegand

he/him

PhD Student in Natural Language Processing @ University St. Gallen (HSG) and Uni-Konstanz

Publications

Conference Papers

A Convexity-Dependent Two-Phase Training Algorithm for Deep Neural Networks

Published in Proceedings of the 17th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K KDIR 2025), 2025

Proposes a two-phase optimization algorithm that detects when the loss landscape transitions from non-convex to convex regions, switching from Adam to Conjugate Gradient to substantially improve convergence speed and accuracy.

Recommended citation: Hrycej, T., Bermeitinger, B., Pavone, M., Wiegand, G.-H., & Handschuh, S. (2025). "A Convexity-Dependent Two-Phase Training Algorithm for Deep Neural Networks." Proceedings of the 17th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management, 78–86.
Download Paper

Leveraging Generative AI for Enhancing Domain-Driven Software Design

Published in Proceedings of the Upper-Rhine Artificial Intelligence Symposium (URAI 2024), 2024

Demonstrates that a 4-bit quantized Code Llama fine-tuned with LoRA on a consumer GPU can generate syntactically correct domain-specific JSON objects for Domain-Driven Design, partially automating metamodel creation.

Recommended citation: Wiegand, G.-H., Stepniak, F., & Baier, P. (2024). "Leveraging Generative AI for Enhancing Domain-Driven Software Design." Proceedings of the Upper-Rhine Artificial Intelligence Symposium, 2024, 41–50.
Download Paper