Two-Way Regularized Logistic Regression with Dynamic Image Regressors

• Activity Number: 108
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): T Siva Tian*+ and Jianhua Z. Huang
• Companies: University of Houston and Texas A&M University
• Keywords: Logistic regression ; Two-way regularization ; Classification ; Spatio-temporal data

Abstract

We propose a novel two-way regularized logistic regression method to classify patients based on the spatio-temporal features presented in the dynamic MEG images. One important characteristic distinguishing the proposed method from existing classification methods is that our method uses a time series of images as predictors. Moreover, we use reduced-rank representation and spatio-temporal regularization to overcome the statistical and computational challenges associated with the high dimensionality of the predictors. A multi-level coordinate descent algorithm is utilized to optimize the log-likelihood function.