Learning Causal Networks Topology From Streaming Graph Signals

Modern data analysis and processing tasks usually involve large sets of data structured by a graph. Typical examples include brain activity supported by neurons, data shared by users of social media, and traffic on transportation or energy networks. There are often settings where the graph is not readily available, and has to be estimated from data. This paper focuses on estimating a network structure capturing the dependencies among streaming graph signals in the form of a possibly directed, weighted adjacency matrix. Several works proposed centralized offline solutions to address this problem, without paying much attention to the distributed nature of networks. We start from a centralized setting and show how, by introducing a simple yet powerful data model, we can infer a graph structure from streaming data with a distributed online learning algorithm. Our algorithm is tested experimentally to illustrate its usefulness, and successfully compared to a centralized offline solution of the literature.