Download presentation

Presentation is loading. Please wait.

Published byFrancisco Hubbart Modified about 1 year ago

1
infoVis 2012 Kasper Dinkla, Michel A. Westenberg, and Jarke J. van Wijk

2
Introduction Related work ◦ Node-Link Based Representations ◦ ZAME: Interactive Large-Scale Graph Visualization TimeMatrix User Study Conclusion

3
Gene Regulatory Network (GRN) ◦ Low in-degree Every gene is regulated by only a few other genes. Therefore, all nodes of the network have a low in- degree. ◦ Scale-free out-degree There are few genes that regulate many others, and many genes that regulate few others. Therefore, the network’s out-degree distribution follows a power law. ◦ Few cycles The network has few cycles because genes rarely (indirectly) regulate each other both ways.

4
node-link diagrams (low edge to node ratio) ◦ high number of intersecting edges adjacency matrices (high edge to node ratio) ◦ space-inefficient, sparse network green: promotion red: inhibition orange: both blue: unspecified

5
Compressed Adjacency Matrices (CAM) ◦ Compactness This enables a detailed overview of the entire network. ◦ Localization of motifs This enables quick detection of subnetworks of interest. ◦ Consistent arrangement This facilitates interaction while preserving visual orientation.

6

7

8

9
the conversion of a network to a CAM is not trivial and consists of six steps: 1.the network is decomposed into weakly connected components 2.nodes with identical neighborhoods are grouped 3.strongly connected components are detected and grouped to form a DAG 4.the nodes of the DAG are partitioned into layers 5.the layers are turned into blocks that form the backbone of the CAM 6.the blocks are concatenated to form a cascade from which node positions

10
directed graph G = (V,E) V : the set of vertices (nodes) E : the set of directed edges between vertices of G

11
G I = (V I,E I ) of G, V I : represent non-overlapping subsets of V with identical neighborhoods E I : the set of directed edges of G I

12

13

14
Layers map directly to blocks ◦ a block B i is derived from its corresponding layer L i H i and V i, that specify the horizontal and vertical ordering of L i ’s vertices in the CAM

15
partitioning the vertices of L i into five classes: Leaf (P L ) ◦ Vertex in L i without successors Short root (P SR ) ◦ Vertex in L i that has a successor but no predecessors and all successors are leaves in L i+1 Long root (P LR ) ◦ Vertex in L i that has a successor but no predecessors and is not a short root Short hub (P SH ) ◦ Vertex in L i that has a predecessor and successor, and all successors are leaves in L i+1 Long hub (P LH ) ◦ Vertex in L i that has a predecessor and successor, but is not a short hub

16

17

18

19

20

21

22

23

24

25

26

27

28
categorizing visual analytic tasks in temporal social network analysis (Tasks 1, 2, and 3), proposing an adjacency-matrix-based visual representation (TimeMatrix) for analyzing temporal graphs that complement node-link temporal graph visualization techniques, and supplementing TimeMatrix with interaction techniques supporting highly interactive visual exploration of real-world social networks across multiple levels of analysis.

29

Similar presentations

© 2016 SlidePlayer.com Inc.

All rights reserved.

Ads by Google