Faculty Recruiting Make a Gift

Modeling Temporal Structures in Dynamic Networks

30 Jan
Monday, 01/30/2017 10:00am to 12:00pm
Lederle Graduate Research Center, Room A311
Ph.D. Dissertation Proposal Defense
Speaker: Kun Tu

"Modeling Temporal Structures in Dynamic Networks"

A dynamic network is a network whose size or structure changes because of the emergence and disappearance of nodes or edges. It is used to model evolving relationships among groups of individuals. Knowledge of those relationships is usually learned by studying the network structure. Considerable work has been conducted on learning network structure, however, due to the complexity of dynamic networks, there is considerable room for improvement to obtain better analysis results. This thesis proposes generative models for the temporal structures of dynamic networks and to characterize their changes. The knowledge of the evolving network structures allows analyses such as missing edge prediction,  and community and lifetime detection. These analyses help to understand real-world complex networks such as social networks, consumer-product networks and contact networks.  In the first part of the thesis, we study bipartite networks constructed from online dating website data, then propose a Latent Dirichlet Allocation (LDA) model to predict edges and apply it to build a recommender system. We then consider more general dynamic networks and present network generative models to detect communities and their lifetime. We present low-rank tensor decomposition technique to learn the generative models. In this work, we discovered that the rank of tensor decomposition plays an important role in retrieving ground truth communities accurately. We study tensor decomposition and propose a Component Augmentation algorithm to reduce the error of learning of the generative models and improve the community detection result. Finally, the last contribution of the thesis focuses on a method based on the network generative models to predict and recover missing data in a dynamic network.

Advisor: Donald F. Towsley