BIMSA > Engineering Mathematics Seminar: Fundamentals and Frontiers in Control, Filtering, State Estimation, and Signal Processing Quantile tensor factor regression with interaction effects and its application to multimodal data analysis
Quantile tensor factor regression with interaction effects and its application to multimodal data analysis
Organizers
Xiaopei Jiao , Jiayi Kang
Speaker
Pengfei Pi
Time
Thursday, May 8, 2025 2:30 PM - 4:00 PM
Venue
A3-2-303
Online
Zoom 435 529 7909 (BIMSA)
Abstract
Multimodal data analysis plays a pivotal role in advancing our comprehension of the brain, contributing significantly to fields such as neuroscience, psychology, psychiatry, and neurology. Multimodal data are consistently modeled as tensor covariates. For numerous clinical and psychological outcomes, quantile regression is highly valued due to its stability and flexibility. This article explores quantile regression involving tensor covariates and applies it to multimodal data analysis. In this proposed approach, tensor covariates are assumed to adhere to a factor structure, from which new feature variables are derived. Subsequently, the main effects of these new feature variables and their interactions are considered in the quantile regression. The article introduces a rapid and efficient method for predicting the dependent variable in quantile regression with interaction effects. The main theoretical results of our approach have been established. The accuracy and stability of the proposed algorithm are validated through extensive simulation experiments. Finally, the proposed method is applied to analyze SKCM and ADHD data, demonstrating superior predictive accuracy and faster computational speed compared to existing methods.
Speaker Intro
Pengfei Pi received the B.S. degree in Mathematics from Sichuan University, Chengdu, China, in 2020. He is currently a Ph.D. candidate in the School of Mathematical Sciences, Shanghai Jiao Tong University. His current research interests include time-varying networks, tensor data analysis, and multi-modal integration.