Network inference from the timing of events in coupled dynamical systems

Area of Science:

• Complex Systems
• Network Science
• Computational Social Science

Background:

• Spreading phenomena, such as opinion formation and disease propagation, are influenced by underlying network structures.
• While network topology's effect on spreading is well-studied, inferring network structure from observed event timings remains a challenge.

Purpose of the Study:

• To investigate the inverse problem of inferring unknown network structures from the timing of events observed at different nodes.
• To evaluate the accuracy of network reconstruction using statistical similarity of event timings.

Main Methods:

• Numerical investigation of two event-based stochastic processes: a generic event propagation model and a variant of the SIRS epidemiological model.
• Analysis of pairwise statistical similarity between event timing sequences using event synchronization and Event Coincidence Analysis (ECA).
• Using functional connectivity (mutual similarity of event sequences) as a proxy for structural connectivity (physical links).

Main Results:

• Both event synchronization and ECA can achieve reasonable accuracy in predicting network structures.
• Sparser networks are generally reconstructed more accurately than denser ones, particularly for larger networks.
• ECA demonstrates superior reconstruction accuracy compared to event synchronization for sparser, larger networks.

Conclusions:

• Event timing analysis, particularly ECA, offers a viable method for inferring network structures from spreading phenomena.
• The accuracy of network reconstruction is dependent on network sparsity and size, with sparser networks yielding better results.