2015/12/07 by Ricardo Pio Monti, Monti, Ricardo Pio, Christoforos Anagnostopoulos +3
Biochemistry, Genetics and Molecular Biology · Neuroscience · #Bioinformatics and Genomic Networks #FOS: Computer and information sciences #Functional Brain Connectivity Studies #Gene expression and cancer classification #Machine Learning (stat.ML)
paper · pdf · doi:10.48550/arxiv.1512.01947
openalex publication_date 2015/12/07 · openalex created_date 2022/09/21 · openalex updated_date 2026/07/28
In neuroimaging data analysis, Gaussian graphical models are often used to\nmodel statistical dependencies across spatially remote brain regions known as\nfunctional connectivity. Typically, data is collected across a cohort of\nsubjects and the scientific objectives consist of estimating population and\nsubject-specific graphical models. A third objective that is often overlooked\ninvolves quantifying inter-subject variability and thus identifying regions or\nsub-networks that demonstrate heterogeneity across subjects. Such information\nis fundamental in order to thoroughly understand the human connectome. We\npropose Mixed Neighborhood Selection in order to simultaneously address the\nthree aforementioned objectives. By recasting covariance selection as a\nneighborhood selection problem we are able to efficiently learn the topology of\neach node. We introduce an additional mixed effect component to neighborhood\nselection in order to simultaneously estimate a graphical model for the\npopulation of subjects as well as for each individual subject. The proposed\nmethod is validated empirically through a series of simulations and applied to\nresting state data for healthy subjects taken from the ABIDE consortium.\n