Many large-scale networked systems such as Online Social Networks (OSNs)are often represented as annotated graphs with various node or link attributes. Such a representation is usually derived from a snapshot that is obtained through measurements. These graph representations enable researchers to characterize the connectivity of these systems using graph analysis. However, captured snapshots of large networked systems are likely to be distorted. Furthermore, commonly-used graph analysis characterizes the connectivity of a graph in an indirect fashion and generally ignores graph dynamics.
This multi-disciplinary research program designs, develops and rigorously evaluates theoretically grounded techniques to accurately measure and properly characterize the connectivity structure of large-scale and dynamic networked systems. More specifically, the project examines various graph sampling techniques for collecting representative samples from large and evolving graphs. It also investigates how multiscale analysis can be used as a powerful technique to characterize the key features of the connectivity structure of large dynamic networked systems at different scales in space and time. The developed techniques will be used to characterize fundamental properties of the friendship and various interaction connectivity structures in different OSNs.
This project promises to identify the underlying technical and social factors that primarily drive the structural properties and dynamic nature of OSN-specific connectivity structures. It will produce new models for friendship and interactions in OSNs, a large archive of anonymized datasets and new tools for OSN measurement, simulation and analysis. The latter will be incorporated into newly-developed courses in Computer Science and Sociology, and will be freely distributed.
