Network Theoretic Analysis of Maximum a Posteriori Detectors for Optimal Input Detection

Abstract

In this paper we characterize the performance of a class of maximum-a-posteriori (MAP) detectors for network systems driven by unknown stochastic inputs, as a function of the location of the sensors and the topology of the network. We consider two scenarios: one in which the changes occurs in the mean of the input, and the other where the changes are allowed to happen in the covariance (or power) of the input. In both the scenarios, to detect the changes, we associate suitable MAP detectors for a given set of sensors, and study its detection performance as function of the network topology, and the graphical distance between the input nodes and the sensors location. When the input and measurement noise follow a Gaussian distribution, we show that, as the number of measurements goes to infinity, the detectors' performance can be studied using the input to output gain of the transfer function of the network system. Using this characterization, we derive conditions under which the detection performance obtained when the sensors are located on a network cut is not worse (resp. not better) than the performance obtained by measuring all nodes of the subnetwork induced by the cut and not containing the input nodes. Our results provide structural insights into the sensor placement from a detection-theoretic viewpoint. Finally, we illustrate our findings via numerical examples.