A1206
Title: Predicting responses from weighted networks with node covariates in an application to neuroimaging
Authors: Daniel Kessler - University of North Carolina at Chapel Hill (United States) [presenting]
Keith Levin - University of Wisconsin (United States)
Liza Levina - University of Michigan (United States)
Abstract: The setting where many networks are observed on a common node-set is considered, and each observation comprises edge weights of a network, covariates observed at each node, and an overall response. The goal is to use the edge weights and node covariates to predict the response while identifying an interpretable set of predictive features. The motivating application is neuroimaging, where edge weights encode functional connectivity measured between brain regions, node covariates encode task activations at each brain region, and the response is disease status or score on a behavioral task. An approach is proposed that constructs feature groups based on assumed community structure (naturally occurring in neuroimaging applications). Two feature grouping schemes that incorporate both edge weights and node covariates are proposed, and algorithms for optimization are derived using an overlapping group LASSO penalty. Empirical results on synthetic data show that our method, relative to competing approaches, has similar or improved prediction error along with superior support recovery, enabling a more interpretable and potentially more accurate understanding of the underlying process. The method is also applied to neuroimaging data from the Human Connectome Project. The approach is widely applicable in neuroimaging, where interpretability is highly desired.
