Learning causal information flow structures in multi-layer networks

We study causal influence structures between the patterns of a multilayer network. Multi-layer networks are networks in which different types of activities between users represent different types of edges, i.e., layers. We measure the causal influence between network patterns via directed information, and investigate how to learn the influence patterns when users can engage in interactions in multiple contexts. We evaluate the proposed methods using both synthetic and real-world datasets, and demonstrate that directed information measures can be utilized to identify the causal relations between network structures.