Characterising the dynamics of unlabelled temporal networks

Annual Talk.



Abstract: Networks form the structural backbone of complex systems, and this backbone often evolves over time, leading to temporal networks. Tools from dynamical systems and time series have been extended to temporal networks by interpreting them as trajectories in a graph space. In this talk, I’ll focus on temporal networks with unlabelled nodes, a scenario that arises in practice due to technical challenges or privacy constraints. Without node labels, there’s no direct correspondence between a network snapshot and a unique adjacency matrix, making it difficult to characterize the dynamics. Using graph invariants, quantifiers like autocorrelation functions and sensitivity to initial conditions can be applied to this unlabelled setting. Through synthetic examples and empirical cases, I’ll show how these methods recover key dynamical properties, even without access to node labels.