Speaker: Sanmi Koyejo, Assistant Professor of Computer Science, University of Illinois at Urbana-Champaign
Abstract: Functional magnetic resonance imaging has transformed the scope and scale of data analysis in neuroscience, providing an unprecedented opportunity to investigate the structure and function of the human brain, discover associations with behavior, and diagnose disorders. With limited samples, meaningful analysis of such high dimensional data must incorporate knowledge gleaned from expertise, experimental evidence and statistical considerations. In particular, restricting the degrees of freedom based on spatial structure has become an important design paradigm, enabling the recovery of parsimonious and interpretable results, and improving storage and prediction efficiency.
I will outline a family of scalable probabilistic machine learning techniques for decoding and exploratory analysis of neuroimaging data, which incorporate the domain knowledge of spatial smoothness and spatial sparsity, leading to improved prediction accuracy and interpretability. These will include techniques for decoding brain state from task data, exploratory analysis via structured probabilistic principal components analysis, and canonical components analysis for joint decomposition of functional, structural and behavioral data. In addition, I will outline new tools for the estimation and analysis of time varying brain connectivity.