科研成果详情

In this paper, we propose an eXtreme deep model with feature interactions and generation for CTR prediction, called xDeepFIG. The feature generation module fully leverages some advantages of convolutional neural network (CNN) to generate new local and global features, and concatenates them with raw features. Such new fully fused features are shared by both the deep neural network (DNN) and compressed interaction network (CIN), which can learn both implicit and explicit high-order feature interactions automatically. Numerical results on two benchmark datasets for CTR demonstrates such feature fusion can bring some advantages and the xDeepFIG outperforms recent baseline models.