Prediction Models Based on Large Scale Image Data

Purdue University, West Lafayette IN

Investigators

Abstract

Research in statistics involves the development and understanding of models based on data. Generally, these data are in the form of numbers, but more recently, statisticians have begun to develop models for data in the form of images. These functional image models have broad applications in neuroscience, engineering, and biomedical practice. This research will further the development of these image models. This project will also include the development of new courses at the undergraduate and graduate levels to train students in the use and understanding of these models. This project is to develop an integrated research program that studies a broad class of large scale functional image models. The PI aims to develop the adaptive and/or local region regression, the finite mixture regression, and the transformation survival regression with ultra-high dimensional image data. The key advantages of these models are to preserve sharp edges for better interpretation, to incorporate the heterogeneity in the population for better representation, and to handle sophisticated censored data. The theoretical contributions of the proposed research are made towards addressing fundamental issues across several disciplines, including nonparametric statistics and machine learning. These functional image models have broad applications in neuroscience, engineering, and biomedical practice. Courses will be developed to train students in the use and understanding of these models.