Pascal Welke

Pascal Welke is a PostDoc Researcher at the Department of Machine Learning, Technical University of Vienna, working on the StruDL project (2023-2027) funded by the Vienna Science and Technology Fund (WWTF). His research focuses on enhancing graph neural networks (GNNs) through theoretical and practical advancements in expressivity, pooling mechanisms, and global feature integration.

• Active in graph representation learning, neural architecture design, and model interpretability
• Collaborates on interdisciplinary projects involving material science and computational linguistics
• Contributor to NeurIPS, ICLR, and ACL conferences

Recent publications analyze the role of expressivity in GNN performance, introduce loop-based WL hierarchies, and develop model distillation techniques. His work spans both theoretical insights (homomorphism-based representations) and practical applications (edge device optimization, nonwoven material analysis). No formal awards or students listed in available records.