The development of Deep Learning models applicable for dynamic graphs respecting the variety of existing graph types and the associated difficulties,

• Graph Neural Networks Designed for Different Graph Types: A Survey
• A Note On The Modeling Power Of Different Graph Types

The consideration of different dynamics w.r.t. the graph structure and attributes and consequent adaptations of established approaches,

• Marked Neural Spatio-Temporal Point Process Involving a Dynamic Graph Neural Network
• Using local activity encoding for dynamic graph pooling in stuctural-dynamic graphs
• Continuous-time generative graph neural network for attributed dynamic graphs

The examination of the expressivity of Deep Learning techniques on graphs,

• Weisfeiler–Lehman goes dynamic: An analysis of the expressive power of Graph Neural Networks for attributed and dynamic graphs

Graph Learning Architectures

• Flow-Attentional Graph Neural Networks

The above work was/is funded through the GAIN and GraphPCBS projects by the Federal Ministry of Research, Technology and Space Germany (BMFTR, formerly BMBF) under funding codes 01IS20047A, according to the 'Policy for the funding of female junior researchers in Artificial Intelligence' and 16ME0877 according to the "KMU-innovativ' guideline. Professor Thomas continues to fulfill her role as PI in this project after transferring to the University of Greifswald.