Bundling: A clutter reduction technique and its application to Alzheimer study
Event date:  October 12, 2017, 2:00 PM  to 3:00 PM

Talk Abstract

Dense and complex data visualizations suffer from occluded items which hinders insight retrieval. This is especially the case very large graph or trails set. To address cluttering issues, several techniques propose to visually simplify the representation, often meeting scalability and computational speed limits. Among them, bundling techniques provide a visual simplification of node-link diagrams by spatially grouping similar items. This presentation strives to bridge the gap between the technical complexity of bundling techniques and the end-point user.

The first aim of this thesis was to improve the understanding of graph and trail bundling techniques as a clutter reduction method for node-link diagrams of large data-set. To do so, we created a data-based taxonomy that organizes bundling methods on the type of data they work on. From this thorough review and based on a formal definition of path bundling, we propose a unified framework that describes the typical steps of bundling algorithms in terms of high-level operations and show how existing methods classes implement these steps. In addition, we propose a description of tasks that bundling aims to address and demonstrate them through a wide set of applications.

Although many techniques exist, handling large data-sets and selectively bundling paths based on attributes is still a challenge. To answer the scalability and computational speed issues of bundling techniques, we propose a new technique which improves both. For this, we shift the bundling process from the image to the spectral space thereby increasing computational limits. We address the later by proposing a streaming scheme allowing bundling of extremely large data-sets.

Finally, as an application domain, we studied how bundling can be used as an efficient visualization technique for societal health challenges. In the context of a national study on Alzheimer disease, we focused our research on the analysis of the mental representation of geographical space for elderly people. We show that using bundling to compare the cognitive maps of dement and non-dement subjects helped neuro-psychologist to formulate new hypotheses on the evolution of Alzheimer disease. These new hypotheses led us to discover a potential marker of the disease years before the actual diagnosis.


Event location:  Universität Stuttgart
Allmandring 19
70569  Stuttgart