Recent Development in Model Selection for Multivariate Nonparametric Regression

Title:  Recent Development in Model Selection for Multivariate Nonparametric Regression

 

Speaker: Prof. Hao Helen Zhang (Department of Mathematics, The University of Arizona)

 

Time: 16:00-17:00pm; June 18; 2013

 

Place: Conference Room A404, Department of Mathematical Sciences

 

AbstractMultivariate nonparametric regression methods are important statistical

analysis tools and widely used in regression, classification, and density estimation. Without presuming the regression function form, nonparametric methods are more flexible than parametric methods and capable of capturing complex nonlinear patterns hidden in data. In the literature, variable and model selection for nonparametric regression methods is a fundamental yet challenging problem due to curse of dimensionality. The difficulty of searching the optimal model within the large model space becomes even more prominent today, due to proliferation of large data sets produced in various scientific fields. Recently there have been some breakthrough works in the area. A variety of cutting-edge model selection techniques have been developed for nonparametric regression. The major advantages of these modern approaches are their tractable theoretical properties, efficient computation, and state-of-art performance in real data analysis. The first half of this talk gives a general review on the recent development and results in model selection for nonparametric regression. The second part will focus on several new regularization methods developed in the context of smoothing spline ANOVA models. Both theoretical and numerical results are presented.

 

ContactYing Yang