Nonparametric Link Prediction for Networks and Bipartite Graphs

• Activity Number: 105 - Inference for Networks and Graph-Structured Data
• Type: Invited
• Date/Time: Monday, August 9, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Jiashen Lu and Kehui Chen*
• Companies: University of Pittsburgh and University of Pittsburgh
• Keywords: network; link prediction; recommender system; missing observations

Abstract

In this talk will discuss a nonparametric link prediction framework for networks and bipartite graphs. In particular, we will discuss how to understand the missing mechanism and to deal with missing observations, when and how to use side information for link prediction, and how to improve the prediction accuracy for new entries (nodes). The proposed statistical framework leads to a simple algorithm with competitive performance, and also provides a new perspective to understand various existing methods for recommender systems in practice.