A scalable matrix-based framework for social dynamics learning
Abstract
The objective of this project is to design a scalable framework for dynamic learning in social networks. The framework will enable robust learning which is resilient to network evolution and provides integrity. Social dynamics will be investigated. Matrix factorization methods will be used to build computational heterogeneous models to discover patterns in real time. Reliability and robustness of the framework will be implemented by integrating heterogeneous data types and filtering out fraudulent information.
Intellectual Merit
The framework will provide solutions to some major existing drawbacks and limitations in social network analysis. The proposed interdisciplinary research will lead to advanced computational models to better understanding social dynamics, conventional sociological theories and assist in developing new ones. Understanding, mining and being able to predict dynamic patterns is interesting for several fields like marketing, security, and Web search. The results can be generalized to biological and technological networks.
Broader Impact
There is an emerging need to systematically investigate the modeling, managing, and mining of large-scale networks and graphs in social networks and biological systems. The results from this project will show the advantages of matrix techniques in solving real-life problems in e-Commerce and social networks.
Use of FutureGrid
FutureGrid will be used to compare performance of algorithms on GPU using CUDA technology and Matlab.
Scale Of Use
run experiments approximately a few times per month.