Download presentation
Presentation is loading. Please wait.
Published byIrma Berry Modified over 9 years ago
1
The FlowVizMenu and Parallel Scatterplot Matrix: Hybrid Multidimensional Visualizations for Network Exploration Christophe Viau, École de technologie supérieure, Montreal Michael J. McGuffin, École de technologie supérieure, Montreal Yves Chiricota, Université du Québec à Chicoutimi, Chicoutimi Igor Jurisica, Ontario Cancer Institute, Toronto
2
Network exploration by graph metrics ?
3
Computed metrics: Degree
4
Network exploration by graph metrics Computed metrics: Degree Closeness centrality Clustering coefficient K-core decomposition...
5
Network exploration by graph metrics Computed metrics: Degree Closeness centrality Clustering coefficient K-core decomposition...
6
Multi-dimensional visualizations Scatterplot Matrix (SPLOM) Parallel Coordinates
7
Related work Using a Scatterplot Matrix (SPLOM) and Node-Link Diagram to visualize a graph GraphDice [Bezerianos et al., 2010]
8
Related work Integration of scatterplots and parallel coordinates Steed et al., 2009 Holten and van Wijk, 2010 Yuan et al., 2009
9
Our proposed interface
10
Parallel Scatterplot Matrix Our proposed interface
11
Parallel Scatterplot Matrix FlowVizMenu Our proposed interface
12
Parallel Scatterplot Matrix FlowVizMenu Attribute-Driven Layout Our proposed interface
13
A sequence of scatterplots
15
Rotating scatterplots around the y-axis
18
Parallel Scatterplot Matrix (P-SPLOM) Scatterplot Matrix (SPLOM) Rotating around x- or y-axes causes a transition from Scatterplot Matrix (SPLOM) to stacked Parallel Coordinates. Parallel Coordinates
19
Ordering of axes within P-SPLOM Problem: traditional SPLOM ordering doesn’t yield useful parallel coordinates. Repeated axes: useless for parallel coordinates Axes are repeated in each row and column
20
Ordering of axes within P-SPLOM Solution: order the axes according to a Latin square. Useful parallel coordinates Each row and column contains each axis once.
21
Scatterplot Staircase (SPLOS) Inspired partly by quilts [Watson et al. 2008] Sequence of scatterplots: treats one dimension differently. Scatterplot Staircase (SPLOS): all dimensions treated uniformly; every adjacent pair of plots share an axis. Parallel coordinates: more difficult to judge correlations than in scatterplots [Li et al., 2010]
22
FlowVizMenu Variant of a FlowMenu with embedded visualization Smoothly animated transitions Brushing and linking More than two dimensions possible with PCA
23
FlowVizMenu In-out gesture to quickly select axes of scatterplot
24
Attribute-Driven Layout (ADL) ADL: Layout based on a scatterplot selected in the FlowVizMenu. The network layout can be a mixture of Attribute-Driven Layout (ADL) Manual layout Force-directed layout
25
Demo
26
Initial user feedback Five bioinformaticians used our prototype and gave feedback. All had experience working with network data. Results: Pros: Exploring along multiple metrics, smooth transitions, and integration of views were judged useful All participants stated they would use the interface if it were made available to them Cons: Some pairings of metrics within the scatterplots may not be useful Too many hotkeys + button combinations in the current prototype
27
Contributions: Three hybrid multidimensional visualization techniques for visualizing networks
28
A Parallel Scatterplot Matrix (P-SPLOM) that transitions between a scatterplot matrix and parallel coordinates
29
Contributions: Three hybrid multidimensional visualization techniques for visualizing networks A Parallel Scatterplot Matrix (P-SPLOM) that transitions between a scatterplot matrix and parallel coordinates
30
Contributions: Three hybrid multidimensional visualization techniques for visualizing networks A Parallel Scatterplot Matrix (P-SPLOM) that transitions between a scatterplot matrix and parallel coordinates A FlowVizMenu to quickly select the dimensions for an embedded scatterplot
31
Contributions: Three hybrid multidimensional visualization techniques for visualizing networks A Parallel Scatterplot Matrix (P-SPLOM) that transitions between a scatterplot matrix and parallel coordinates A FlowVizMenu to quickly select the dimensions for an embedded scatterplot
32
Contributions: Three hybrid multidimensional visualization techniques for visualizing networks A Parallel Scatterplot Matrix (P-SPLOM) that transitions between a scatterplot matrix and parallel coordinates A FlowVizMenu to quickly select the dimensions for an embedded scatterplot An Attribute-Driven Layout to configure the graph according to a scatterplot of graph metrics
33
Future directions Application to other domains Evaluation of performance and usability Exploration of the design space of each visualization (e.g., on a small screen)
34
Acknowledgments We thank our collaborators for their feedback: SAP Business Objects Members of Jurisica Lab at OCI Members of the Multimedia Lab at ETS This research was funded by an SAP Business Objects ARC Fellowship, NSERC, and the FQRNT.
35
Thank you
37
P-SPLOM: variants
38
P-SPLOM: Pearson correlation coefficient
39
P-SPLOM: Latin square
40
P-SPLOM: another latin square
41
Scatterplot Staircase
42
Parallel Scatterplot Matrix (P-SPLOM) Rotating around x- or y-axes causes a transition from Scatterplot Matrix (SPLOM) to stacked Parallel Coordinates
43
Ordering of axes within P-SPLOM The traditional SPLOM ordering doesn’t produce interesting parallel coordinates Repeated axes: useless for parallel coordinates
44
P-SPLOM ordering We explored variants of latin square
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.