User Tools

Site Tools


projects:year4:15.4

15.4 - Graph Sampling, Summarization, and Touch-Based Visual Analytics for Large Complex Systems

Project - Summary

We seek to enable interactive visual analytics of large-scale graphs using novel graph sampling methods and touch-based interfaces. Recently, there is a significant interest in modeling and studying real-world complex systems as large-scale graphs with numerous interconnected or interacting entities. Many real-world systems such as online social networks (OSN), world wide web (WWW) and Internet topology maps (ITMs) are very large, so capturing them in their entirety, analyzing them to extract useful information, and visualizing them for decision making are resource-consuming and challenging tasks. It is necessary to develop graph sampling approaches and integrate them with human-computer interfaces to study these large-scale graphs to understand the underlying real-world systems. These networks are large and decentralized which make the global structure of them invisible. One of the approaches to overcome this issue is sampling. In sampling small subsets of nodes and links from a network are selected. Sampling makes it possible to study a small part of the networks while preserving features of the original network. There are many related works to “graph sampling”. However, they target different types of networks context and they have different characteristics. It is necessary to develop graph sampling approaches and integrate them with human-computer interfaces to study these large-scale graphs to understand the underlying real-world systems.

The PIs will investigate sources of information loss in a graph sampling process and identify fundamental factors that need to be carefully considered in a sampling design. We also plan to develop a software system as an extension to open source libraries (igraph/networkX/boost) that employs different sampling methods to estimate important graph characteristics. Developing an extension to igraph/networkX/boost, instead of a standalone application, allows more seamless integration of our work with other CVDI projects. Furthermore, the project will improve interactive visual analysis of large graphs by prototyping interface methods in combination with machine analytics. We will develop multitouch and gesture techniques to provide intuitive user control of navigation, filtering, clustering, and highlighting during visual analysis. The efficiency and clarity of interfaces is critical for the success of visual analytics systems and helps users understand results and analysis processes.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Chrisoph Borst PI Not available Not available UL Lafayette
Mehmet Engin Tozal PI Not available Not available UL Lafayette
Nicholas Lipari Student Not available Not available UL Lafayette
Maryam Heidari Student Not available Not available UL Lafayette

Project - Impact and Uses/Benefits

CVDI members use graphs to model, analyze and visualize their underlying systems and business processes to make informed decisions and gain insights from complex heterogeneous data. Effective graph sampling, visualization, and interaction methods for various tasks with complement ongoing CVDI projects that employ graphs as a tool to model and analyze real-world systems and complex data. Novel approaches and methods may lead to advantages in monitoring operations, understanding data, retaining and acquiring customers, increasing revenue, etc. We aim to deliver potentially faster and more powerful visual analytic interfaces and more accessible and understandable graph browsing interfaces.

Project - Deep Dive

Project - Documents

FilenameFilesizeLast modified
15.4_year_4_presentation.pdf2.2 MiB2019/08/22 11:50
15.4_year_4_executive_summary.pdf163.1 KiB2019/08/22 11:50
15.4_year_4_ip_letter_combined.pdf371.0 KiB2019/08/22 11:50
15.4_year_4_final_report.pdf985.6 KiB2019/08/22 10:33
projects/year4/15.4.txt · Last modified: 2019/08/22 10:34 by sally.johnson