Unfolding Edges

Interactive visualization of multivariate edge attributes.

Network visualization displays the relationships between data elements. Each node represents a data element (e.g. people, devices, groups) and edges(e.g. friendship, Facebook connection, communication) indicate the relationships between items. Network visualization helps viewers to comprehend the structure of complex relationships between elements. This technique is widely used as a way of representing networking, bioinformatics, software engineering, database and machine learning.

Multivariate network visualization is widely used to show both the structure of the network and its properties. The layout does not only contain nodes and their relationships but also attributes of nodes and links. For example, if Facebook connections are presented in multivariate network visualization, viewers can see the overall formation of relationships and also explore several characteristics of every friend(node) such as age, name, the duration of connection and etc.

To effectively visualize multiple attributes, it is common to use on-edge encoding utilizing multiple visual variables. Several examples are depicted in in the figure above. They include line width for edge weight, color coding for edge type, fuzziness for uncertainty, arrows and gradients for edge direction, and oscillating waves to represent edge length. Despite using this variety of encoding, the spatial issue of edges makes it challenging to read edge attributes in practice. Especially, if there are too many nodes in the network graph, understanding the overall structure of visualization becomes very difficult because of edge crossings and node overlaps.

Researchers have presented several approaches to address these challenges. For example, they integrated detail enhancements in the context of node attributes or they exploited different graph visualization types such as adjacency matrices. Meanwhile, studies for enhancing edge attribute details by using interactions were somewhat limited. Edge interaction allows custom selection and filtering, so it improves the legibility of visualization and also minimizes occlusions.

In a EuroVis 2021 paper, M. J. Bludau and team propose an approach called “unfolding edges” using both interactions and visual encoding. By inducing viewers an interaction of unfolding edge, visualization presents multiple attributes. Showing additional attribute details on the same view by using interactions and transitions improves the readability of the visualization. Viewers would not lose the global context and openly explore nodes and edges.

The figures above are two exemplifications of using on-demand edge unfolding. When viewers select a certain edge, it turns into a more detailed visualization. Semantic zooming creates additional space by blurring unrelated edges and nodes. Additional representation in this space helps viewers to understand multiple attributes while keeping the global context at the same time.

The upper figure shows a form of edge fanning. Various curves of unfolded sub-edges involve multiple source types (color-coded), uncertainty(dashed line), and ordering and label presenting the time dimension of an edge. The lower figure uses rotation to place a selected edge in a horizontal position. A rectangular space containing color-coded sequential data visualization shows when the selected edge is unfolded. Unfolding edge by interactions such as zooming and rotation allows viewers to explore multiple attributes of the network presented in the visualization.

Depending on the data and the field, edge unfolding has many potential use cases. For example, it would be applicable to presenting historical network analysis. Connections between historical figures may be based on historical records such as messages or letters. By presenting various types of attributes such as duration, direction, and date in unfolded edge space, visualization would present observation and interpretation of network structure among historical actors.

“Unfolding edges” is applicable to visualizing complex multivariate edge attributes in a scalable fashion. Although this methodology needs proper evaluation, it may provide a more sophisticated way to present attribute detail compared to global edge encoding approaches. Finding answers for these questions: “How to promote edge interaction?”, “How to indicate hidden detail”, “How to expend on-demand enhancement of edge details for global use” may be helpful to propose a broad way to create space for interactive edge visualization.

Related posts:

Ameerpet Technologies Provides the Best Full Stack Java Training in Hyderabad & Online With 100% placements. Learn Full Stack Java Course from Basics to advanced and get real- time experience. If you…

PlayDapp Global Media Manager, Ryan Chown joined AMA Lovers Club for a great session, and to help spread the word about Along with the Gods The (Along With The Gods) AWTG_PlayDapp PROJECT team was…

It goes without saying that this sport has many adoring fans from all over the world. It is after all the highest class of international auto racing, which only consists of twenty drivers, from ten…