B0570
Title: Fully symmetric graphical lasso for dependent data
Authors: Saverio Ranciati - Universita di Bologna (Italy) [presenting]
Alberto Roverato - University of Padova (Italy)
Alessandra Luati - Imperial College London (United Kingdom)
Abstract: A method is proposed to analyze multivariate data with intrinsic symmetrical structures and, in general, to solve problems belonging to the class of dependent samples inference, such as case-control studies, matched and paired data. To this aim, we propose the fully symmetric graphical lasso, a penalized likelihood method with a fused type penalty function that takes into explicit account the natural symmetrical structure within and between symmetrical blocks of the data (or samples). The implementation leverages an alternating directions method of multipliers algorithm to solve the corresponding convex optimization problem. The procedure is applied to various real-world datasets, concerning air pollution and brain fMRI scans.
