Dissertation Defense by Jiashen Lu

Thursday, April 28th at 10:30 AM

Wesley W. Posvar Hall, Department of Statistics, Seminar Room

Title: Nonparametric Predictions for Network Links and Recommendation Systems

Abstract: In this work, we develop methodologies to make nonparametric predictions in relational data. Prominent examples of relational data include user-user network interactions and user-item recommendation systems. For social networks, we follow a new latent position framework and develop prediction methods in pure cold-start scenarios where new nodes do not have any observed links to start with. For recommendation systems, we first develop a zero-imputation method to address the challenges of heterogeneous missing, and then make predictions for unobserved values and for new users or items. We provide theoretical guarantees of the proposed method and demonstrate its good performance in real data analysis as well as simulations.

Committee Chair and Advisor: Dr. Kehui Chen