Investigating scalable visual metaphors for very large networks

• Grant number: 15/14426-5
• Support type: Scholarships abroad - Research Internship - Post-doctor
• Effective date (Start): October 19, 2015
• Effective date (End): May 18, 2016
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal researcher: Maria Cristina Ferreira de Oliveira
• Grantee: Moussa Reda Mansour
• Supervisor abroad: John C. Hart
• Home Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Research place: University of Illinois at Urbana-Champaign, United States

Abstract

This project is presented as part of a Research Internship Application (BEPE) for the Computer Graphics@Illinois, Illinois University at Urbana-Champaign, under the supervision of Professor John C. Hart. The ongoing postdoctoral project investigates novel scalable visual metaphors for abstract data visualization, aiming at developing new paradigms to facilitate understanding, exploration and user-driven mining of large data sets. This internship project is focused on novel and scalable visual metaphors for large social networks. Social network data typically carry attribute information associated to the individuals and to their relationships. Different approaches have been introduced to extract and identify information of interest in social networks, and community identification is one of them.Some methods focus on identifying groups of individuals (communities) based on their relationships, while others try to group individuals based on the common information they share. Integrating both approaches is not straightforward, as different mathematical and computational methods must be implemented and integrated into a unified framework. Moreover, many methods are not applicable due the high computational effort required. In this project our goal is to develop a new method based on parallel programming strategies to identify underlying communities in a network based on highlighting the information shared by their components. Our solution relies on a single unified mathematical methodology. As a proof-of-concept, we intend to apply the proposed method to scientific co-authorship networks extracted from the well-known Lattes Platform made available by CNPq, the Brazilian national science funding agency. (AU)