Research (Large) Network Visualisation Tools

[CharlotFranssen]

• Network with a lot of nodes, edges
• Create summary of findings

[WouterMulder9]

A good visualization is difficult to define, as it largely depends on the given context. Based on research, 4 key features of a good visualization are formulated as:

1. Minimum edges intersection.
2. Adjacent vertices are closer to each other than not adjacent.
3. Communities are grouped into clusters.
4. Minimum overlapping edges and nodes.

These criteria become more complex to achieve as the size of the network increases. Research focusses on tackling these issues through dimension reduction, feature-based layouts or graph embeddings such as Node2Vec, Verse, and Graph Convolutions.

However, even than networks and graphs can be too large to visualize. Fortunately, there are tools that could help. From these, the most important consideration are:

1. Igraph: 30000 to 50000 nodes; we have experience, but has a poor Python API.
2. GraphViz: A CLI tool, but hard to tune parameters without interactivity.
3. Gephi: Powerful and simple, but abandoned by the developers.
4. Cosmograph (or cosmo on GitHub): Clean interface, very powerful. Mainly in browser.

Sources:

• https://cosmograph.app/
• https://towardsdatascience.com/large-graph-visualization-tools-and-approaches-2b8758a1cd59

[WouterMulder9]

First part seems to be based on research articles, but they are not mentioned. Other than that, looks good!