Learning the Wireless Interference Graph via Local Probes

Due to the densification and ambitious spectral efficiency targets, wireless networks are becoming increasingly interference-limited. Effectively managing the interference and accordingly allocating communication resources strongly hinge on knowing the interference signals. Acquiring such information, however, is often infeasible as networks’ scale and complexity grow. This paper considers a multiuser decentralized wireless network proposes an algorithm for learning the interference signals by aggregating the minimal data collected by individual users. Specifically, each user has a binary quantization of the interference level it experiences from other users. The proposed algorithm aggregates the local binary data to form an estimate of the interference network. Sufficient conditions for an optimal inference are delineated, and the proposed algorithm is demonstrated to enjoy certain optimality guarantees.