Visualizing Networks Large & Small

Vahan Yoghourdjian

Under the supervision of

Tim Dwyer, Kim Marriott, Michael Wybrow, Karsten Klein

What?

High-quality layout

Large networks

Small networks

Structural summary

What?

• Can we create better network diagrams?

Why?

• Existing algorithms are complex and inflexible

• We need:• Application specific semantics

• Grid points

• Variable node sizes

• Flow, etc.

• Enhanced readability• Compactness

• Proximity of connected nodes

• Non-overlap, etc.Yoghourdjian, Vahan, et al. "High-quality ultra-compact grid layout of grouped networks." IEEE

transactions on visualization and computer graphics 22.1 (2016): 339-348.

How?

• Is it practical to use constraint programming to solve network layout problems?

groups contain only respective nodesHow?

• Decision variables• Node placement• Edge routing

• Constraints

• Objective function• Manhattan stress• Group compactness• Grid compactness

Add additional constraints

u v

u and v should not overlap

How?

• Will these techniques produce high-quality layouts in reasonable time?

• Experimental study with 3 common solvers• 5 minutes timeout

• Up to 60 nodes

• Grid size• Total edge length• Crossings

10x11334

9x8294

7x7250

Force-Directed with Grid Snap

Large Neighborhood

Search

Optimal

What?

• Big data – Large networks

• “Overview first, zoom and filter, detail on demand” Shneiderman

• How can we use small clear diagrams to visualize very large networks?

Why?

• Graph Browsing

Bach, Benjamin, et al.

"Small MultiPiles: Piling

time to explore temporal

patterns in dynamic

networks." Computer

Graphics Forum. Vol. 34. No. 3. 2015.

Why?

• Visualizing Graph Metrics (B-Matrix, GraphPrism,

Brandes and Pich, VoG,

etc.)• Not Scalable

• Detailed Analysis

Kairam, Sanjay, et al. "GraphPrism: compact visualization of network structure." Proceedings of the International Working Conference on Advanced Visual Interfaces. ACM, 2012.

How?

• How to use small diagrams to reveal structural information of large and dense networks?

• Graph Thumbnails• K-Connectivity & K-Core

• Precise

• Canonical

• Scalable

Number of EdgesNumber of NodesNode Degrees

Graph Thumbnail Node-Link Diagram

1-connected

2-connected

3-connected

Graph Thumbnail Node-Link Diagram

Graph 9Graph 2 Graph 3 Graph 4 Graph 5 Graph 6 Graph 7 Graph 8 Graph 10Graph 1

Similarity Study

• 5-star rating

• User feedback

Similarity Study

• 5-star rating

• User feedback

Similarity Study

• 5-star rating

• User feedback

Analyzing the Results

GT NL MX dense graphs

clustered

sparse disconnected

Analyzing the Results

GT

dense graphs

clustered

sparse disconnected

Analyzing the Results

NL

dense graphs

clustered

sparse disconnected

Analyzing the Results

MX

dense graphs

clustered

sparse disconnected

Future Work

• Analyzing the results• Compare similarity rating of graphs with different

structure

• Compare similarity rating of graphs with similar structure

• Training affect

• Check for positive or negative connotation in feedback

GT

NL

MX

Future Work

• Matching Study • Ask participants to match

Graph Thumbnails to node-link diagrams

• Vice versa

Example: given a graph thumbnail and 3 node-link diagrams

Which node-link diagram matches the graph thumbnail?

Future Work

• Connectivity detection• 1-connected, 2-connected subsets

• Graph Thumbnails, node-link diagrams

& adjacency matrices

How many 1-connected subsets can you find in

this diagram?• 3• 5• 10

Example: given a graph thumbnail

Questions

• Graph thumbnail evaluation• Number of edges• Node degrees• Stars

• Consider alternate representations• Icicle Plot

• Other existing tools or methods that serve similar objectives

Questions

• Other applications for graph thumbnails• Dynamic networks

• Animation• Blur / sharpness

• Uncertainty• Fuzziness / blur

• Analyzing the results of the similarity study

• Other studies that I should conduct• Similarity study with larger graphs and smaller thumbnails• Interview experts / practitioners• Compare community detection algorithms with Graph Thumbnails

Questions

• Focus-Context visualization

Thank you

For more information or contact, visit Vahany.com

Some parts of the work described today is found in “High-Quality Ultra-Compact Grid Layout of Grouped Networks”