Title: Adaptive regularization with applications to brain imaging
Authors: Jaroslaw Harezlak - Indiana University School of Public Health-Bloomington (United States) [presenting]
Abstract: The problem of adaptive incorporation of multi-modal brain imagining data sources in multiple linear regression settings is addressed. In the presented example, we model scalar outcome dependence on the brain cortical properties, e.g. cortical thickness and cortical area. We utilize both connectivity and spatial proximity information to build adaptive penalty terms in the regularized regression problem. The general idea of incorporating external information in the regularization approach via linear mixed model representation has been recently established in our prior proposal named ridgefield Partially Empirical Eigenvectors for Regression (riPEER). We incorporate multiple sources of information, including structural and functional connectivity network structure as well as the spatial distance between the cortical regions to estimate the regression parameters with multiple penalty terms via a riPEER extension called AIMER (Adaptive Information Merging Estimator for Regression). We present a simulation study testing various realistic scenarios and apply AIMER to data arising from the Human Connectome Project (HCP) study.