Network Topology Identification using PCA and its Graph Theoretic Interpretations

We solve the problem of identifying (reconstructing) network topology from steady state network measurements. Concretely, given only a data matrix X where the X_{ij} entry corresponds to flow in edge i in configuration (steady-state) j, we wish to find a network structure for which flow conservation is obeyed at all the nodes. This models many network problems involving conserved quantities like water, power, and metabolic networks. We show that identification is equivalent to learning a model A_n which captures the approximate linear relationships between the different variables comprising X (i.e. of the form A_n X \approx 0) such that A_n is full rank (highest possible) and consistent with a network node-edge incidence structure. The problem is solved through a sequence of steps like estimating approximate linear relationships using Principal Component Analysis, obtaining f-cut-sets from these approximate relationships, and graph realization from f-cut-sets (or equivalently f-circuits). Each step and the overall process is polynomial time. The method is illustrated by identifying topology of a water distribution network. We also study the extent of identifiability from steady-state data.