|
Activity Number:
|
482
- Statistical Methods in the Analysis of High-Order Structural Data with Possible Structural Changes
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
|
Abstract #300338
|
Presentation
|
|
Title:
|
Tensor Regression and Imaging-Based Inference
|
|
Author(s):
|
Heping Zhang* and Long Feng and Xuan Bi
|
|
Companies:
|
Yale University and Yale University and University of Minnesota
|
|
Keywords:
|
brain image;
tensor regression;
total variation;
shrinkage;
eigenvalue;
cognition
|
|
Abstract:
|
The use of brain-imaging data to predict cognitive disabilities has drawn increasing attentions in both psychology and public health. As brain-imaging data are usually represented by three or even higher order tensors, we aim to develop a class of high-order tensor (imaging) regression models to address this issue. A key novelty of our method is that it takes advantages of properties of imaging coefficients in the form of high-order tensors. This is achieved by an innovative shrinkage method that plays the role of total variation for high order tensors. Theoretically, we simultaneously provide the computational and statistical errors of the proposed estimates under a restricted eigenvalue condition and certain initialization requirements. Numerically, we examine simulated data as well as analyze the data from the Philadelphia Neurodevelopmental Cohort. Our findings demonstrate the superiority of our models in unraveling the underlying data structures and relations in imaging-genetic data.
|
Authors who are presenting talks have a * after their name.
