Bayesian Topology Learning and noise removal from network data

Abstract Learning the topology of a graph from available data is great interest in many emerging applications. Some examples are social networks, internet things networks (intelligent IoT and industrial IoT), biological connection sensor traffic network patterns. In this paper, inference approach proposed to learn underlying structure given set noisy multi-variate observations, which modeled as signals generated Gaussian Markov Random Field (GMRF) process. A factor analysis model applied represent latent space where basis related structure. An optimal filter also developed recover observations. final step, an optimization problem recovered signals. Moreover, fast algorithm employing proximal point method has been solve efficiently. Experimental results both synthetic real show effectiveness recovering inferring graph.