This article is the second in a series of wrap-ups from VizWorld reader T.J. Jankun-Kelly (@dr_tj) of the EuroVis2010 conference currently underway.  If you missed it, see Part 1.

Day of EuroVis was the first day of parallel tracks. Today, your humble scribe attended the A track, which largely focused on Information Visualization; for the impression of some of the B track, see Charl Botha’s twitter feed (@cpbotha) or the general EuroVis twitter list (via @eagereyes).

Session 3A: Graph Visualization

SmallWorlds: Visualizing Social Recommendations; Brynjar Gretarsson, John O’Donovan, Svetlin Bostandjiev, Christopher Hall, Tobias Hollerer

Gretarsson, part of the group at UCSB, presented a graph-based recommendation system integrated with the Facebook API. Unlike other recommendation systems, the system works entirely via positive recommendations (“likes”) of a limited subset of a population (a user’s friends, as opposed to all Facebook users). These limitations in comparison to other systems are addressed via visualization and interaction: Users can see which friends have similar likes, change the weight of their friend’s recommendations, and also change the weight of their own likes all based upon simple, web-based interactions. The rendering is done server-side and can be seen on the SmallWorlds Facebook App page.

See the rest of TJ’s writeup after the break.

Image-Based Edge Bundles: Simplified Visualization of Large Graph; Alexandru Telea, Ozan Ersoy

Edge Bundles are a popular method to reduce visual clutter in graph visualization; however, they tend to cause major node crossings to be somewhat ambiguous. Telea & Ersoy address this using an image-based splatting of the automatically clustered edges. Edges that go in the same direction are grouped together into a single shape skeleton which is then used as the basis for the edge splats. Interaction methods to clear up ambiguity in cluster crossings are also introduced.

Winding Roads: Routing edges into bundles; Antoine Lambert, Romain Bourqui, David Auber

Another edge bundling work, Lambert et al. from LaBri, focus on arbitrary graph rendering in the plane as opposed to the more traditional circular graphs. A hybrid quadtree & voronoi decomposition of the nodes is used to gather and route edges into bundles using an iterated shortest path calculation. Alternate edge rendering methods, including a edge splatting like the previous work, are explored.

GraphDice: A System for Exploring Multivariate Social Networks; Anastasia Bezerianos, Fanny Chevalier, Pierre Dragicevic, Niklas Elmqvist, Jean- Daniel Fekete

Extending Elmqvist’s Scatter Dice work from InfoVis 2008, GraphDice use a matrix of scatterplots of graph statistics to explore complex graphs. Chevalier presented how novel interactions and animation allow for effective stepping between comparing graph dimensions. A history based interaction schema was also discussed.

Session 4A: Space & Color

Space-in-time and time-in-space self-organizing maps for exploring spatiotemporal patterns; Gennady Andrienko, Natalia Andrienko, Sebastian Bremm, Tobias Schreck, Tatiana von Landesberger, Peter Bak, Daniel Keim

Gennady Andrienko led this discussion of two different clustering methods for understanding spatially-aware time varying data such as cell phone usage patterns. One clustered spatial data by their patterns in time; the other one clustered time patterns via their similarity in spatial information. For example, one may group areas with similar call patterns over time or group times that have similar call areas by the location where the call was made. Embedded visualization of the related spaces (for time) or the related times (for space) was also presented.

Alleviating the Modifiable Areal Unit Problem within Probe-Based Geospatial Analyse; Thomas Butkiewicz, Remco Chang, Zachary Wartell, William Ribarsky

Butkievicz, from the VisCenter at UNCC, discussed a problem with the selection and visualization of spatial data: The Modifiable Areal Unit Problem. In geospatial data, arbitrary spatial selections often are biased in the measures associated with that data: The selection may cross boundary lines, may include regions of vastly different density and thus skew the statistics, and so on. Butkievicz outlined several tools for automatically detecting and correcting for these anomalies via density aware region correction algorithms. Several applications in urban planning were discusses as motivation.

Real-Time Temporal-Coherent Color Contrast Enhancement for Dichromats; Gustavo Mello Machado, Manuel M. Oliveira

The color portion of the Space & Color session, Oliveira discussed a method to dynamically correct for color vision deficiency in visualization. They specifically address methods to ensure maximum hue contrast while approximately maintaining the luminance difference. The approach is space and time coherent and is implemented via a GPU. Several demonstrations were given for continuous/scientific visualization-like images while discrete/information visualization images were presented in the paper.

Session 5A: Navigating Abstract Spaces

Dynamic Multi-View Exploration of Shape Spaces; Stef Busking, Charl Botha, Frits Post

A collection of models can be examined in shape space in order to determine their similarities and find outliers which can be useful in multiple domains, such as medicine and modeling. In this presentation, Busking outline a three part coordinated-view visualization of shape space for such purposes. First, a star-coordinates like view of shape space is used to provide an overview, allowing quick interaction to ease out dimensions of interest. Next, selection in the overview space can be made to create interpolations of the shape which are displayed in the second window; by “scrubbing” the overview window, a quick visual evolution is displayed. Finally, these evolutions can be further explored via a collection of shape projections over the shape path in the third window; color is used to indicate areas of significant difference in the shapes along the path. Source for the project is available on Google.

HyperMoVal: Interactive Visual Validation of Regression Models for Real-Time Simulation; Harald Piringer, Wolfgang Berger, Jürgen Krasser

A project growing out of industrial collaboration with VRViz, HyperMoVal uses multiple coordinated views to quickly explore and validate regression models in complex engineering models. The multiple projections are used to be able to detect outliers in the high-dimensional space and to refit the models based upon new selections. The complex visualization, while powerful, did meet with some resistance from users due to deviations in workflow from their original system.

Generating and Exploring a Collection of Topological Landscapes for Visualization of Scalar- Valued Functions / William Harvey, Yusu Wang

Harvey & Wang from OSU presented a novel method for summarizing the topological characteristics of high dimensional data via creating a topological map between the space and a height field over a treemap. Using the contour-tree decomposition of Weber et al., Harvey & Wang characterize the extrema in the high-dimensional field, translating saddle and merge points into splits in the treemap. Arbitrary remapping of the chosen initial floor of the data field was also presented.

Session 6A: Applications

An Interactive Visual Analytics System for Bridge Management; Xiaoyu Wang, Wenwen Dou, Shen-En Chen, William Ribarsky, Remco Chang

Another paper from the VisCenter at UNCC, this presentation discussed a bridge analytics system integrating much of the work from Charlotte over the past several  years. Image and video analytics are combined with data from professional bridge inspectors to facilitate the management of bridges an bridge repair resources.

Pathline: A Tool For Comparative Functional Genomics; Miriah Meyer, Bang Wong, Mark Styczynski, Tamara Munzner, Hanspeter Pfister

Meyer and her collaborators discuss another visually aesthetically pleasing and effective design for comparative functional genomics visualization. Using the insight that unrolled paths in a tangled biological network can still provide effective insight, Pathline linearizes such networks and provides a small-multiples display of the temporal behavior of genes and metabolites across species for pathways of interest. Several examples from comparative genomics were provides, and the source code is available at pathline.org

Video Visualization for Snooker Skill Training; Markus Höferlin, Edward Grundy, Rita Borgo, Daniel Weiskopf, Min Chen, Iwan W. Griffiths, W. Griffiths

The final talk of the day was a video analytics system for snooker training from Swansea & Stuttgart. Video from multiple cameras were used to record snooker practice; information from these frames are extracted and visualized over space and time. This allows, with professional feedback, a potentially more effective means to understand rookie behavior. Interest in the application prototype from professionals has been noted.