Abstract

In many scientific and engineering applications, predictors are naturally grouped, for example, in biological applications where assayed genes or proteins can be grouped by biological roles or biological pathways. When the group structures are available among predictors, people are usually interested in identifying both important groups and important variables within the selected groups. Among existing successful group variable selection methods, some methods fail to conduct the within group selection. Some methods are able to conduct both group and within group selection, but the corresponding objective function is non-convex, which may require extra numerical effort. In this talk, we will present a convex penalty for both group and within group variable selection. We develop an efficient group-level coordinate descent algorithm for solving the corresponding optimization problem. We also study the non-asymptotic properties of the estimates in the high-dimensional setting, where the number of predictors can be much larger than the sample size. Numerical results indicate that the proposed method works well in terms of both variable selection and prediction accuracy. We also apply the proposed method to American Cancer Society Breast Cancer Survivor dataset.

Tuesday, January 15, 2013

3:00– 4:00 P.M.

372 Science and Engineering Building

(Refreshments at 2:30-3:00 PM in the kitchen area adjacent to 368 SEB)