Explore a cutting-edge statistical method for estimating latent graphical models from multimodal functional data in this USC Probability and Statistics Seminar talk. Delve into the innovative approach developed to address the gap in principled statistical methods for analyzing joint multimodal functional data, particularly in neurological and biological sciences. Learn about the generative perspective used to model the data generation process and the identification of transformation operators for mapping from observation to latent space. Discover how the proposed estimator simultaneously estimates transformation operators and the latent graph through functional neighborhood regression. Gain insights into the application of this method to analyzing simultaneously acquired multimodal brain imaging data for understanding underlying brain functional connectivity. The talk also covers joint work with Katherine Tsai, Boxin Zhao, and Sanmi Koyejo, providing a comprehensive overview of this advanced statistical technique and its potential impact on multimodal data analysis in various scientific fields.
Latent Multimodal Functional Graphical Model Estimation
USC Probability and Statistics Seminar via YouTube
Overview
Syllabus
Mladen Kolar: Latent multimodal functional graphical model estimation (U Chicago)
Taught by
USC Probability and Statistics Seminar