Trees and Networks CS 7250 S PRING 2020 Prof. Cody Dunne N - - PowerPoint PPT Presentation

▶
trees and networks
SMART_READER_LITE
LIVE PREVIEW

Trees and Networks CS 7250 S PRING 2020 Prof. Cody Dunne N - - PowerPoint PPT Presentation

Apr 09, 2023 288 likes •846 views

Trees and Networks CS 7250 S PRING 2020 Prof. Cody Dunne N ORTHEASTERN U NIVERSITY Slides and inspiration from Michelle Borkin, Krzysztof Gajos, Hanspeter Pfister, 1 Miriah Meyer, Jonathan Schwabish, and David Sprague B URNING Q UESTIONS ? 2 P

slide-1
SLIDE 1

Trees and Networks

CS 7250 SPRING 2020

  • Prof. Cody Dunne

NORTHEASTERN UNIVERSITY

1

Slides and inspiration from Michelle Borkin, Krzysztof Gajos, Hanspeter Pfister, Miriah Meyer, Jonathan Schwabish, and David Sprague

slide-2
SLIDE 2

BURNING QUESTIONS?

2

slide-3
SLIDE 3

PREVIOUSLY, ON CS 7250…

3

slide-4
SLIDE 4

4

“Overview first, zoom and filter, and details on demand.”

  • Ben Shneiderman

“The Shneiderman Mantra”

Shneiderman, 1996

slide-5
SLIDE 5

5

Interaction

  • Complexity reduction
  • Static = specific story told to you, versus interactive =

viewer discovers the story

  • Enables data exploration, insight, reasoning for oneself
  • Makes it personal to the viewer
  • Dive deeper!

Why interaction?

slide-6
SLIDE 6

6

Interaction

  • Interaction requires human time and attention
  • Human-guided search vs. Automatic feature

detection vs. Interactive visualizations

  • Find balance between automation and relying on

the human in the loop to detect patterns A few footnotes...

Based on Slide by Hanspeter Pfister

slide-7
SLIDE 7

NOW, ON CS 7250…

7

slide-8
SLIDE 8

TREES & (MAINLY) NETWORKS

8

slide-9
SLIDE 9

GOALS FOR TODAY

  • Learn the definition of a network (including node, edge)
  • Learn the definition of a tree
  • Learn common visual encoding techniques for network

data (i.e., node-link diagram, adjacency matrix), and the advantages of each one.

9

slide-10
SLIDE 10

Hall of Fame or Hall of Shame

10

slide-11
SLIDE 11

11

Sudhahar et al., 2015

US presidential election network for 2012 primaries.

  • Nodes: key entities from noun phrases.

Sized by degree.

  • Edges: relationships from verbs.

Colored by positive (green) and negative (red) weights.

slide-12
SLIDE 12

12

Sudhahar et al., 2015

slide-13
SLIDE 13

13

Andrew Bergman, 2014

slide-14
SLIDE 14

14

slide-15
SLIDE 15

15

Tree = undirected, connected, acyclic network Network = entities and relationships between them

(vertex, entity) (edge, tie, relationship) (graphs)

slide-16
SLIDE 16

16

Modified from slide by Frank van Ham

Note all the same network, just different layouts!

Networks

  • A network G consists of a set of nodes N and a set of edges E
  • An edge en1,n2 ∈ E connects two nodes n1, n2 ∈ N
  • E.g., G = {1,2,3,4}, E = {(1,2),(1,3), (2,3),(3,4),(4,1)}
slide-17
SLIDE 17

Isolate Main connected component

A bunch of definitions

17

Modified from slide by Frank van Ham

slide-18
SLIDE 18

18

“Treemap”

slide-19
SLIDE 19

19

  • Primary concern is the spatial layout of nodes and

edges, a.k.a. graph drawing

  • The goal is often to effectively depict the graph

structure for topology-based tasks:

  • connectivity, path-following
  • network distance
  • clustering
  • ordering (e.g., hierarchy level)
  • But not always topology-based tasks. E.g.,

understanding attributes, statistics, metrics

Slide based on Miriah Meyer

slide-20
SLIDE 20

Quantitative Tasks

Mackinlay, 1986

20

Spatial Layout

slide-21
SLIDE 21

21

Spatial Layout

slide-22
SLIDE 22

22

Cleveland & McGill, 1984 Heer & Bostock (2010)

Spatial Layout

slide-23
SLIDE 23

Flickr Query for “Mouse”

slide-24
SLIDE 24

Tweets of the #Win09 Workshop

slide-25
SLIDE 25
slide-26
SLIDE 26
slide-27
SLIDE 27

27

http://londonist.com/2016/05/the-history-of-the-tube-map

slide-28
SLIDE 28

28

http://news-explorer.mybluemix.net

slide-29
SLIDE 29

29

Node-Link Visualizations - Marks & Channels

Node Edge Gestalt Principles: Grouping, Proximity, Connectedness Color Size Shape Color Thickness Style Direction

slide-30
SLIDE 30

30

  • Nodes are distributed in space,

connected by straight or curved lines

  • Typical approach is to use 2D space to

break apart breadth and depth

  • Often space is used to communicate

hierarchical orientation

Slide based on Miriah Meyer

Node-Link Visualizations

slide-31
SLIDE 31

31

Pros:

  • understandable visual mapping
  • can show overall structure, clusters, paths
  • flexible, many variations

Cons:

  • automatic layout algorithm deficiencies
  • time consuming to run
  • non-deterministic results
  • heuristics with sometimes poor results
  • not good for dense graphs - hairball problem!

Node-Link Visualizations

Slide based on Miriah Meyer

slide-32
SLIDE 32

Mike Bostock

slide-33
SLIDE 33

Mike Bostock

slide-34
SLIDE 34

34

https://observablehq.com/@d3/force-directed-graph

Layout Algorithm: D3 Force-Directed

slide-35
SLIDE 35

35

Kobourov, 2012

Force-Directed Layout Algorithms

slide-36
SLIDE 36

36

Upcode, 2020

Dashboard of the COVID-19 Virus Outbreak in Singapore

2020.01.21–03.12

slide-37
SLIDE 37

37

Dashboard of the COVID-19 Virus Outbreak in Singapore

2020.01.21–03.12

Upcode, 2020

slide-38
SLIDE 38

In-Class Curation — Network Planarity Party

38

~25 min

slide-39
SLIDE 39
slide-40
SLIDE 40
slide-41
SLIDE 41

Hachul & Jünger, 2006 Graph A Graph B

Layout Algorithm Comparisons

slide-42
SLIDE 42

Sugiyama, 2002, p. 14 User performance Huang et al., 2007, etc. Simple rules or heuristics Davidson & Harel, 1996 Global and local readability metrics Purchase et al., 2002 Dunne et al., 2015

How to compare?

slide-43
SLIDE 43

43

  • Quickly run out of space!
  • Tree breadth often grows exponentially
  • Layout algorithms are slow and heuristics
  • Solutions:
  • scrolling or panning
  • filtering or zooming
  • aggregation & simplification

Scale Problems...

Slide based on Miriah Meyer

slide-44
SLIDE 44

http://www.yasiv.com/graphs#HB/blckhole

slide-45
SLIDE 45

45

https://gephi.org/

slide-46
SLIDE 46

46

“Treemap”

slide-47
SLIDE 47

47

Slide based on Miriah Meyer

Alternate to node-link visualization for dense & weighted networks

slide-48
SLIDE 48

48

Adjacency Matrix

Henry & Fekete (2006)

slide-49
SLIDE 49

49

Pros:

  • great for dense graphs
  • visually scalable
  • can spot clusters

Cons:

  • row order affects what you can see
  • abstract visualization
  • hard to follow paths

Slide by Miriah Meyer

slide-50
SLIDE 50

50

https://bost.ocks.org/mike/miserables/

slide-51
SLIDE 51
slide-52
SLIDE 52

52

http://higlass.io/

slide-53
SLIDE 53

Henry & Fekete, 2007

53

MatLink

slide-54
SLIDE 54

Henry et al, 2007

54

NodeTrix

slide-55
SLIDE 55

Yang et al., 2016; Demo

55

MapTrix

https://vimeo.com/278433529 https://vimeo.com/182970812