A0957
Title: Reduced-rank finite mixture regression for multivariate response via low-rank regularization
Authors: Suyeon Kang - University of Central Florida (United States) [presenting]
Kun Chen - University of Connecticut (United States)
Weixin Yao - UC Riverside (United States)
Abstract: Given the rapid growth in data volume and access to diverse data sources, data complexity and heterogeneity have escalated across many fields. The aim is to extend reduced-rank estimation to mixture modeling by proposing a new class of reduced-rank multivariate mixture regression models. These models handle multiple continuous responses under population heterogeneity while extracting low-dimensional structure. Computationally efficient EM-type algorithms that incorporate both a rank penalty and an adaptive nuclear-norm penalty are derived, enabling simultaneous subgroup identification, parameter estimation, and rank selection. The monotonicity of the penalized likelihood sequence and the asymptotic consistency of the estimators are proven. Simulation studies and real data analysis have been carried out to validate the effectiveness and practical usefulness of the proposed methods. The R package rrMixture is developed for the implementation and is publicly available on CRAN.
