Dynamic Vertex Replacement Grammars
This work addresses the limitation of existing graph grammars in handling temporal dynamics, which is important for applications involving evolving relational data such as social networks or biological systems.
The authors tackled the problem of modeling time-changing phenomena in relational data by introducing dynamic vertex-replacement grammars (DyVeRG), which extend graph grammars to capture temporal changes and enable forecasting through a novel dyvergence score, demonstrating faithful generation and interpretability on real-world dynamic graphs.
Context-free graph grammars have shown a remarkable ability to model structures in real-world relational data. However, graph grammars lack the ability to capture time-changing phenomena since the left-to-right transitions of a production rule do not represent temporal change. In the present work, we describe dynamic vertex-replacement grammars (DyVeRG), which generalize vertex replacement grammars in the time domain by providing a formal framework for updating a learned graph grammar in accordance with modifications to its underlying data. We show that DyVeRG grammars can be learned from, and used to generate, real-world dynamic graphs faithfully while remaining human-interpretable. We also demonstrate their ability to forecast by computing dyvergence scores, a novel graph similarity measurement exposed by this framework.