Graph Computing and Social Cognitive Analytics for Network Science

When: Wednesday, December 11, 2013 at 1:00 pm
Where: DA 5th fl
Speaker: Ching-Yung Lin, PhD
Organization: IBM Lead, Social and Cognitive Network Research Center, IBM T.J. Watson Research Center
Sponsor: CCNR and Barabási Lab Seminar

Many real-world data are linked. Entities are dependent and thus form large graphs. However, traditional IT software and computer architectures are mainly designed to handle independent data.  Processing, storing, analyzing, retrieving, and visualizing connected data has been a major challenge for Big Data. Novel graph computing technologies with fundamental paradigm shifts are needed in order to advance Network Science.

I am going to introduce our work of “System G”, a complete software stack for Graph Computing, including Graph Database, Middleware for Hardware Optimization, Analytics Library, and Visualization Suites. Graphs may be large or small, static or dynamic, topological or semantic, and property-oriented or Bayesian. In some real world computations, e.g., graph traversal, System G native store may be 100x faster to retrieve data than using traditional database. I will also introduce our application research works on (1) Cognitive Analytics, which utilizes graphical models to understand and predict people’s behavior for Insider Threat Detection or Commerce, (2) Social Analytics, which analyzes collective behaviors of people in social media, and (3) Brain Analytics, which models and visualizes neuron’s dynamic networks of animal brain such as mouse.