A Convex Formulation for Learning Scale-Free Networks via Submodular Relaxation

Part of Advances in Neural Information Processing Systems 25 (NIPS 2012)

Bibtex Metadata Paper Supplemental

Authors

Aaron Defazio, Tibério Caetano

Abstract

A key problem in statistics and machine learning is the determination of network structure from data. We consider the case where the structure of the graph to be reconstructed is known to be scale-free. We show that in such cases it is natural to formulate structured sparsity inducing priors using submodular functions, and we use their Lovasz extension to obtain a convex relaxation. For tractable classes such as Gaussian graphical models, this leads to a convex optimization problem that can be efficiently solved. We show that our method results in an improvement in the accuracy of reconstructed networks for synthetic data. We also show how our prior encourages scale-free reconstructions on a bioinfomatics dataset.