Title: High dimensional classification with artificial neural network
Authors: Taps Maiti - Michigan State University (United States) [presenting]
Abstract: High-dimensional models with correlated predictors are commonly seen in practice. Most proposed statistical models works well either in the low-dimensional correlated case, or in the high-dimensional independent case. Few methods deals with high-dimensional correlated predictors. Neural networks have been applied in practice for years, which have a good performance in correlated predictors due to the non-linearity brought by the activation functions. However, it may have too many parameters in high-dimensional case. With regularization, we are able to apply the neural network to high-dimensional correlated predictors case and obtain a parsimonious model with fairly good theoretical and numerical performance.