Sparse Regularization Methods for Nonparametric Function Estimation

• Activity Number: 220
• Type: Invited
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Leonard Stefanski and Yichao Wu*
• Companies: North Carolina State University and North Carolina State University
• Keywords: attenuation ; convolution ; deconvolution ; noise addition ; variable selection

Abstract

Deconvolution is difficult because of the unfortunate effectiveness of noise in lessening information, usually in the guise of attenuated signals. A corollary is that convolution, or noise addition, can be used as an indicator of the presence of signal. In the context of classifying explanatory variables as important or unimportant, noise added to an unimportant variable has negligible effect on the information in a data set; whereas noise added to an informative variable generally reduces information. Starting from the noise-contamination idea of indicating variable informativeness, the I will describe how the theory, methods, and algorithms from the field of measurement error modeling can be used to develop new methods of variable selection applicable across the full spectrum of model- and algorithmic-based prediction methods. Instances of the general strategy are described for certain prediction methods.