*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title : Learning, Inference, and Optimization of High-Dimensional Asynchronous Event Data
Nan Du
School of Computational Science and Engineering College of Computing Georgia Institute of Technology
Date : Friday, December 04, 2015
Time : 10:00 AM to 12:00 PM EST
Location : KACB 1315
Committee
-------------
Dr. Le Song (Advisor), School of Computational Science and Engineering, Georgia Institute of Technology Dr. Hongyuan Zha, School of Computational Science and Engineering, Georgia Institute of Technology Dr. Yu (Jeffrey) Hu, Scheller College of Business, Georgia Institute of Technology Dr. Christos Faloutsos, School of Computer Science, Carnegie Mellon University Dr. Evgeniy Gabrilovich, Senior Staff Research Scientist, Google Research
Abstract
-------------
The increasing availability and granularity of temporal and spatial event data induced from user activities in on-line media, social networks and medical informatics provide new opportunities and challenges to model and understand user behaviors. In addition to study the macroscopic patterns on the population level, such type of data further enable us to investigate user interactions in a more fine-grained scale to address the "who will do what by when and where ?" question with new exploratory and predictive models. On the other hand, because these myriads of microscopic event data, such as publishing a post, forwarding a tweet, purchasing a product, checking in a place, often arise asynchronously and interdependently, they require new representing and analyzing methods far beyond those based on independent and identically distributed data models. In this dissertation, we present a novel probabilistic framework for modeling and reasoning about repeated sequences of event data. Within the proposed framework, we introduce a pipeline of newly developed statistical models, state-of-the-arts learning algorithms and a general purpose software library for multivariate point process to tackle several canonical problems in practice, including : (1) latent network structure recovery, (2) continuous-time influence estimation and maximization, (3) time-sensitive recommendation and (4) temporal document clustering, in the field of social network analysis, recommendation systems, and document modeling.
