Distributed hyperplane learning using consensus algorithm for sensor networks

In this paper, we study distributed classification of targets in a large scale sensor network setting. Specifically, we consider sensor nodes which can measure only a part of the feature vector and whose communication capabilities are limited to only their neighbouring nodes. We formulate a distributed classification algorithm that learns the optimal (large-margin) hyperplane separating the two classes, using the projected-gradient approach. The sensor nodes reach a consensus on the gradient to be used for weight update at each step of the optimization algorithm. We prove the convergence of the proposed algorithm and provide a joint calibration and signature learning method for acoustic sensor networks as an application.