Research Abstracts Online
January - December 2011
University of Minnesota Twin Cities
College of Liberal Arts
PI: Yuhong Yang
High-Dimensional Statistical Problems
These researchers are interested in two problems using high-dimensional data: estimation of classification accuracy and regression estimation. For the first problem, they are exploring the best use of cross-validation so that the resulting confidence interval of classification test error is optimal in some proper measures. The second problem investigates both model selection and model combining methods to improve regression estimation accuracy. Both problems involve intensive computations to search over many possible choices to reach the best or near best solution that balances approximation error, estimation error, and error due to searching.
Gang Cheng, Graduate Student