Title: Parallel Gibbs variable selection for high-dimensional generalized linear models
Authors: Guangbao Guo - Shandong University of Technology (China) [presenting]
Abstract: A novel parallel Gibbs variable selection procedure is proposed for high-dimensional generalized linear models. In the procedure, the data are randomly split into some subsets according to the given rules. We propose a series of weights to obtain optimized stationary distributions. Through the Gibbs method, we can quickly select effective parallel group variables. In aspect of theoretical properties, we obtain convergence of the method.