Project: Interactive and Visual Analysis of Networks

Acronym IVAN (Reference Number: ANR-17-CHR2-0003)
Duration 01/01/2018 - 31/12/2020
Project Topic Our main goal is to create a visual analysis system for the exploration of dynamic or time-dependent networks (from small to large scale). Our contributions will be in three principle areas: (A) novel algorithms for network clustering that are based on graph harmonic analysis and level-of-detail methods; (B) the development of novel similarity measures for networks and network clusters for the purpose of comparing multiple network clusterings and the grouping (clustering) of different network clusterings; and (C) a system for user-driven analysis of network clusterings supported by novel visual encodings and interaction techniques suitable for exploring dynamic networks and their clusterings in the presence of uncertainties due to noise and uncontrolled variations of network properties. Our aim is to make these novel algorithms accessible to a broad range of users and researchers to enable reliable and informed decisions based on the network analysis. A focus in all three areas will be on the incorporation of uncertainty into the analysis and visual encoding to enhance the trust in the decision making. While we are aiming to create tools for a variety of use cases, we specifically focus on two application areas -- social networks such as Twitter as well as brain functional networks. These are two applications where the consortium has a lot of expertise, yet which are very different in terms of users and tasks. Hence, we hope to be able to generalize from these two specific applications. Our team consists of three distinct research labs with expertise in harmonic analysis (Dr. Van De Ville, EPFL, Switzerland), expertise in network visualization (Dr. Fekete, Inria France) and expertise in supporting of visual model building and analysis under uncertainty (Dr. Möller, Uni Wien, Austria). This is a unique combination of skills that is indispensable to successfully tackle the challenges of this endeavour made possible only under the unique requirements of this funding call.
Project Results
(after finalisation)
Describe the expected progress beyond state of the art and more generally the targeted outcomes. They should be clear, measurable, realistic and achievable within the duration of the project. Provide quantitative information when possible. To maximize the impact, we will adhere to principles of open science and make all source code publicly available using github repositories, including example datasets such that main results can be reproduced by other research groups. In particular, we will use available academic git-platforms such as for scientific code co-creation, curation, sharing and testing. The core algorithms of our framework will be provided to network scientists as a library that supports all developments. This library will then be used in our interactive visual analysis framework for network clustering and will incorporate: ● novel ways of comparing different clustering approaches of dynamic networks; ● different metrics and algorithms for meta-clustering of network clusters; ● new visual encodings for uncertainty in network clusters. In addition, we will provide a typology of network exploration tasks based on several user studies on network algorithm developers and domain experts for social networks and brain networks. These two lines of implementations not only allow method developers to test, adapt, and further extend the framework, but also let end-users benefit from the application of the framework to their data in a user-friendly environment.
Call CHIST-ERA Call 2017

Project partner

Number Name Role Country
1 University of Vienna Coordinator Austria
2 National Institute of Research in Computer Science and Automation - Saclay Partner France
3 Federal Institute of Technology in Lausanne Partner Switzerland