Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 7: Poster
SOE 7.1: Poster
Monday, March 18, 2024, 18:00–20:30, Poster D
Bayesian Detection of Mesoscale Structures in Pathway Data on Graphs — •Vincenzo Perri1, Luka V. Petrović2, and Ingo Scholtes1 — 1University of Würzburg, Würzburg, Germany — 2University of Zürich, Zürich, Switzerland
Mesoscale structures are an integral part of the abstraction and analysis of complex systems. For example, they can represent communities in social or citation networks, roles in corporate interactions, or core-periphery structures in transportation networks. Many methods to detect mesoscale structures are founded on the assumption that the interactions are independent. However, when complex systems are dynamic, the dynamics of interactions can invalidate this assumption and jeopardize the analysis. In this work, we derive a Bayesian approach that simultaneously models the optimal partitioning of nodes in groups and the optimal higher-order network dynamics between the groups. This model can capture mesoscale structures with higher-order patterns, which allows us to coarse grain pathway data, and analyse them at the level of groups of nodes. In synthetic data, we show that our method can recover both standard static patterns and dynamic patterns invisible to baselines. In empirical data, we find higher-order patterns at group levels, underlying the practical importance of our method. Finally, we demonstrate the interpretability of our method in an application on a language corpus and detection of vowel-consonant dynamics. From a broader perspective, our approach allows interpretable analysis of a complex system and facilitates our understanding of its structural and temporal patterns.
Keywords: Network Dynamics; Sequential Data; Bayesian Statistics; Mesoscale Structures; Markov Chain Modelling