Path-Based Conditions for the Identifiability of Non-additive Nonlinear Networks with Full Measurements

We analyze the identifiability of nonlinear networks with non necessarily additive node dynamics, where the influence of in-neighbors is represented by a multivariate nonlinear function that cannot necessarily be separated into individual edge functions.

We consider the full measurement case (all the nodes are measured) and we introduce the notion of generic identifiability.

Based on a generic nonlinear matrix associated with an unfolded digraph constructed from the network, we characterize the space of functions that satisfies the generic property.

For directed acyclic graphs (DAGs) composed of analytic functions, we derive a sufficient condition for identifiability based on vertex-disjoint paths from excited nodes to the in-neighbors of each node in the network.

Furthermore, for the class of polynomial functions, by using well-known results on algebraic varieties, we prove that the identifiability is impossible if the vertex-disjoint path condition is not satisfied.

Finally, we show that this identifiability condition is not necessary for the additive nonlinear model, where the node function can be decomposed into a sum of edge-wise nonlinearities.