SEMI-BLIND INFERENCE OF TOPOLOGIES AND SIGNALS OVER GRAPHS
Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis · 2018
Network science provides valuable insights across numerous disciplines including sociology, biology, neuroscience and engineering. A task of major practical importance in these application domains is inferring the network topology from noisy observations over a limited subset of nodes. This work presents a novel approach for joint inference of the network topology and estimation of graph signals from partial nodal observations based on structural equation models (SEMs). SEMs have well-documented merits in identifying the directed topology of complex graphs by capturing causal relationships among nodes. The resultant algorithm iterates between inferring a directed graph that “best” fits the data, and estimating the graph signals over the learned graph. Numerical tests with synthetic as well as real data corroborate the effectiveness of the joint inference approach.