Learning graphical models with hubs

摘要: We consider the problem of learning a high-dimensional graphical model in which there are few hub nodes that densely-connected to many other nodes. Many authors have studied use an l1 penalty order learn sparse graph setting. However, implicitly assumes each edge is equally likely and independent all edges. propose general framework accommodate more realistic networks with nodes, using convex formulation involves row-column overlap norm penalty. apply this three widely-used probabilistic models: Gaussian model, covariance binary Ising model. An alternating direction method multipliers algorithm used solve corresponding optimization problems. On synthetic data, we demonstrate our proposed outperforms competitors do not explicitly illustrate proposal on webpage data set gene expression set.