On the Suitability of Connectivity-Extended Local Embedding for Drawing Multivariate Graphs

Loading...
Thumbnail Image
Date
2014
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
Multivariate networks are present in various domains such as biology, or social science. In such networks, the nodes often have several quantitative attributes, which determine similarity of nodes (e.g., person's characteristics in social networks). When interpreting these networks, often both node connectivity and node similarity need to be analyzed simultaneously. Such analysis can be supported by suitable layouts. We present and evaluate a layout for graphs with multivariate numeric attributes, which combines graph structure and node similarity. It extends local dimension reduction techniques (esp. LLE, MEU, or ISOMAP) with graph connectivity information encoded in techniques' local neighborhood function. We evaluate these extensions and available layouts using two conflicting criteria: distance preservation and graph aesthetics. Although the results vary across data sets, the new approach is able to find a balance of these criteria.
Description

        
@inproceedings{
:10.2312/vmv.20141285
, booktitle = {
Vision, Modeling & Visualization
}, editor = {
Jan Bender and Arjan Kuijper and Tatiana von Landesberger and Holger Theisel and Philipp Urban
}, title = {{
On the Suitability of Connectivity-Extended Local Embedding for Drawing Multivariate Graphs
}}, author = {
Fahnenschreiber, Sebastian
and
Laux, Melvin
and
Landesberger, Tatiana von
}, year = {
2014
}, publisher = {
The Eurographics Association
}, ISBN = {
978-3-905674-74-3
}, DOI = {
/10.2312/vmv.20141285
} }
Citation
Collections