Network visualization by david shelley
This presentation is the property of its rightful owner.
Sponsored Links
1 / 52

Network Visualization by David Shelley PowerPoint PPT Presentation


  • 107 Views
  • Uploaded on
  • Presentation posted in: General

Network Visualization by David Shelley. Some slides adapted from Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com. Outline. Why visualize large networks? 2. Issues when Graphing Large Networks Production Issues Layout Issues

Download Presentation

Network Visualization by David Shelley

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Network visualization by david shelley

Network Visualizationby David Shelley

Some slides adapted from Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Outline

Outline

  • Why visualize large networks?

    2. Issues when Graphing Large Networks

    Production Issues

    Layout Issues

    3. Common solutions to graphing large networks

    4. Conclude with common tools


Outline1

Outline

  • Why visualize large networks?

    2. Issues when Graphing Large Networks

    Production Issues

    Layout Issues

    3. Common solutions to graphing large networks


Why visualize large networks

Why visualize large networks?

“There is nothing better than a picture for making you think of questions you had forgotten to ask (even mentally)” Tukey and Tueky, 1985

“Finding ways to visualize datasets can be as important as ways to analyze them.” Ripley 2005

“Data visualization is good for data cleaning, for exploring data, for identifying treads and clusters, for spotting local patterns, for evaluating modelling output and for presenting resutls. Visualization is essential for Exploratory Data Analysis.” Unwin et. Al.

Quotes found in Graphics of Large Datasets, Visualizing a Million by Unwin et. Al.


Why visualize large networks1

Why visualize large networks?

  • Discover anomalies in the data


Why visualize large networks2

Why visualize large networks?

  • Understand the flow of a network

    Metro Network of Washington DCInternet Service Providers


Why visualize large networks3

Why visualize large networks?

Map of Springfield by Jerry Lerma and Terry Hogan

Understand the relation between geographical objects

How do I get to Moe’s ?


Why visualize large networks4

Why visualize large networks?

  • Use it to find socioeconomic patters.

    19811992

    http://www.mpi-fg-koeln.mpg.de/~lk/netvis/trade/WorldTrade.html


Why visualize large networks5

Why visualize large networks?

  • Conclusion:

    Discover anomalies.

    Understand the flow of a network.

    Understand the relation between geographical objects.

    Use it to find socioeconomic patters.

    Many other reasons not mentioned.


Outline2

Outline

  • Why visualize large networks?

    2. Issues when Graphing Large Networks

    Production Issues

    Layout Issues

    3. Common solutions to graphing large networks


Outline3

Outline

  • Why visualize large networks?

    2. Issues when Graphing Large Networks

    Production Issues

    Layout Issues

    3. Common solutions to graphing large networks


Issues when graphing large networks production issues

Issues when Graphing Large NetworksPRODUCTION ISSUES

Storage

Hard disk space.

RAM (memory).

File formats of data.

Google’s First Production Server

It is not publically known but Wikipedia estimates that

Google maintains over 450,000 servers.

Source: http://flickr.com/photos/jurvetson/157722937/

Graphics of large Datasets Visualizing a Million (Antony Unwin, Martin Theus Heike Hofmann)


Issues when graphing large networks production issues1

Issues when Graphing Large NetworksPRODUCTION ISSUES

Quality

The larger the network, the higher possibility of errors in the data.

Complexity (meaning many not Big-O)

This is a major problem with large networks. More variables, more detail, more categories.

Speed

Currently we are interested in getting results from our graph fast enough to be considered interactive.

Analysis

What algorithms are used. What order of complexity is required for the algorithms.

Graphics of large Datasets Visualizing a Million (Antony Unwin, Martin Theus Heike Hofmann)


Issues when graphing large networks production issues2

Issues when Graphing Large NetworksPRODUCTION ISSUES

Display

The more nodes there are the more pixels on the screen you will need.

The more information that needs to be presented on the screen the more window design and window management become increasingly important.


Issues when graphing large networks production issues3

Issues when Graphing Large NetworksPRODUCTION ISSUES

Display

800 x 600 = 480,000 pixels

1024 x 768 = 786,432 pixels

1920 x 1200 = 2,304,000 pixels

Not enough pixels to display all the nodes!!!


Issues when graphing large networks production issues4

Issues when Graphing Large NetworksPRODUCTION ISSUES

  • Conclusion of production issues:

    Physical Memory Issues.

    Quality of Data Issues.

    Complexity of each element in the graph (not talking about Big-O).

    Speed of loading and handling all the elements

    Analyzing the large data set. Finder better algorithms.

    Display overload. Not enough pixels on a single screen.


Outline4

Outline

  • Why visualize large networks?

    2. Issues when Graphing Large Networks

    Production Issues

    Layout Issues

    3. Common solutions to graphing large networks


Issues when graphing large networks layout issues

Issues when Graphing Large NetworksLAYOUT ISSUES

How to represent an edge?

Labels on Edges

Thickness of Edges

A

Color of Edge

Shape of Edges

Directed Edges


Issues when graphing large networks layout issues1

Issues when Graphing Large NetworksLAYOUT ISSUES

The problem with edges is they can occlude other parts of the graph!!!

Before drawing edges After drawing edges


Issues when graphing large networks layout issues2

Issues when Graphing Large NetworksLAYOUT ISSUES

How to represent a node?

Shapes of Nodes

Size of Nodes

Color of Nodes

A

Labels of Nodes

Location of Nodes

B


Issues when graphing large networks

Issues when Graphing Large Networks

Conclusion:

Production Issues

Storage

Quality

Complexity

Speed

Analysis

Layout Issues

How to represent and edge

How to represent a node


Outline5

Outline

  • Why visualize large networks?

    2. Issues when Graphing Large Networks

    Production Issues

    Layout Issues

    3. Common solutions to graphing large networks


Common solutions to graphing large networks

Common solutions to graphing large networks

Draw important objects on top of other objects.

Notice how the nodes have been covered up by edges.


Common solutions to graphing large networks1

Common solutions to graphing large networks

Aesthetic Considerations

Minimize lines crossing.

Non-overlapping.

Scale edge lengths.

VS

VS

VS


Common solutions to graphing large networks2

Common solutions to graphing large networks

Aesthetic Considerations


Common solutions to graphing large networks3

Common solutions to graphing large networks

Use alpha-blending


Common solutions to graphing large networks4

Common solutions to graphing large networks

Layout Algorithms

Planar layout

Tree layout

Circular/Spiral

And there’s more…


Common solutions to graphing large networks5

Common solutions to graphing large networks

Layout Algorithms Dynamic Networks

Kamada-Kawai (KK) (spring embedder)

Fruchterman-Reingold (FR) Force

And there’s more…

Force Layout Methods such as the Spring Model

http://java.sun.com/applets/jdk/1.4/demo/applets/GraphLayout/example1.html

Java Universal Network/Graph Framework (JUNG) http://jung.sourceforge.net/applet/showlayouts.html


Common solutions to graphing large networks6

Common solutions to graphing large networks

User Interaction

Theus (1996) and Unwin (1999) have proposed there are three broad components of interaction for statistical graphics.

1. Querying

2. Selection and linking

3. Varying plot characteristics


Common solutions to graphing large networks7

Common solutions to graphing large networks

User Interaction

Theus (1996) and Unwin (1999) have proposed there are three broad components of interaction for statistical graphics.

1. Querying

http://adn.blam.be/springfield/


Common solutions to graphing large networks8

Common solutions to graphing large networks

User Interaction

Theus (1996) and Unwin (1999) have proposed there are three broad components of interaction for statistical graphics.

1. Querying

2. Selection and linking

Select View

Stats View

AT&T

Sprint


Common solutions to graphing large networks9

Common solutions to graphing large networks

User Interaction

Theus (1996) and Unwin (1999) have proposed there are three broad components of interaction for statistical graphics.

1. Querying

2. Selection and linking


Common solutions to graphing large networks10

Common solutions to graphing large networks

User Interaction

Theus (1996) and Unwin (1999) have proposed there are three broad components of interaction for statistical graphics.

1. Querying

2. Selection and linking

3. Varying plot characteristics

Sort View

Spanish View

AT e T

AT&T

Sprint

Esprint


Common solutions to graphing large networks11

Common solutions to graphing large networks

User Interaction

Focus + Context

  • Basic Idea:

    • Show selected regions of interest in greater detail (focus)

    • Preserve global view at reduced detail (context)

    • NO occlusion

      • All information is visible simultaneously

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks12

Common solutions to graphing large networks

User Interaction

Focus + Context

Alternative Names for Focus + Context

  • Fisheye views

  • Fisheye lens

  • Continuously variable zoom

  • Nonlinear magnification

  • Hyperbolic views

  • Distortion viewing

  • Rubber sheet views

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks13

Common solutions to graphing large networks

User Interaction

Focus + Context

Applications for Focus + Context

  • Visualization of Networks/Graphs

  • Viewing text

  • Image/Document viewing

  • Cartography

  • Cluster Visualization

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks14

Common solutions to graphing large networks

User Interaction

Focus + Context

Applications for Focus + Context

  • Visualization of Networks/Graphs

  • Viewing text

  • Image/Document viewing

  • Cartography

  • Cluster Visualization

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks15

Common solutions to graphing large networks

User Interaction

Focus + Context

Applications for Focus + Context

  • Visualization of Networks/Graphs

  • Viewing text

  • Image/Document viewing

  • Cartography

  • Cluster Visualization

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks16

Common solutions to graphing large networks

User Interaction

Focus + Context

Applications for Focus + Context

  • Visualization of Networks/Graphs

  • Viewing text

  • Image/Document viewing

  • Cartography

  • Cluster Visualization

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks17

Common solutions to graphing large networks

User Interaction

Focus + Context

Applications for Focus + Context

  • Visualization of Networks/Graphs

  • Viewing text

  • Image/Document viewing

  • Cartography

  • Cluster Visualization

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks18

Common solutions to graphing large networks

User Interaction

Focus + Context

Types of Focus + Context

  • Spatial

    • One Dimensional

      • Easy to apply and understand

    • Two Dimensional

      • Most common, operating on 2D layouts of information

    • Three Dimensional

      • Less common

  • Logical

    • Effect applies to logical structure of the information

  • Combined Spatial and Logical

  • Data Driven Magnification

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks19

Common solutions to graphing large networks

User Interaction

Focus + Context

Types of Focus + Context

  • Spatial

    • One Dimensional

      • Easy to apply and understand

    • Two Dimensional

      • Most common, operating on 2D layouts of information

    • Three Dimensional

      • Less common

  • Logical

    • Effect applies to logical structure of the information

  • Combined Spatial and Logical

  • Data Driven Magnification

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks20

Common solutions to graphing large networks

User Interaction

Focus + Context

Types of Focus + Context

  • Spatial

    • One Dimensional

      • Easy to apply and understand

    • Two Dimensional

      • Most common, operating on 2D layouts of information

    • Three Dimensional

      • Less common

  • Logical

    • Effect applies to logical structure of the information

  • Combined Spatial and Logical

  • Data Driven Magnification

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks21

Common solutions to graphing large networks

User Interaction

Focus + Context

Types of Focus + Context

  • Spatial

    • One Dimensional

      • Easy to apply and understand

    • Two Dimensional

      • Most common, operating on 2D layouts of information

    • Three Dimensional

      • Less common

  • Logical

    • Effect applies to logical structure of the information

  • Combined Spatial and Logical

  • Data Driven Magnification

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks22

Common solutions to graphing large networks

User Interaction

Focus + Context

Types of Focus + Context

  • Spatial

    • One Dimensional

      • Easy to apply and understand

    • Two Dimensional

      • Most common, operating on 2D layouts of information

    • Three Dimensional

      • Less common

  • Logical

    • Effect applies to logical structure of the information

  • Combined Spatial and Logical

  • Data Driven Magnification

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks23

Common solutions to graphing large networks

User Interaction

Focus + Context

Types of Focus + Context

  • Spatial

    • One Dimensional

      • Easy to apply and understand

    • Two Dimensional

      • Most common, operating on 2D layouts of information

    • Three Dimensional

      • Less common

  • Logical

    • Effect applies to logical structure of the information

  • Combined Spatial and Logical

  • Data Driven Magnification

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Example of moiregraph

Example of MoireGraph

http://www.cse.msstate.edu/~tjk/publications/papers/tjk-infovis03.pdf


Common solutions to graphing large networks24

Common solutions to graphing large networks

User Interaction

Focus + Context Limitations

  • Limited degree of magnification?

    • 10X Maximum?

    • Open research question

  • Disorientation

    • Complex transformations might cause viewer to get lost

    • Need effective visual cues to avoid this

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks25

Common solutions to graphing large networks

User Interaction

Focus + Context Strengths

  • Mirrors the way the visual cortex is designed

  • Good navigation tool for interactively exploring data

    • probe regions of interest before committing to navigating to them (easily reversible)

  • Can be combined with other viewing paradigms such as Pan and Zoom

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Common solutions to graphing large networks26

Common solutions to graphing large networks

User Interaction

Focus + Context Alternatives

  • Pan&Zoom

    • Scales to high factors

    • Navigation can be a problem

  • Multiple views at different scales

    • No distortion between scales

    • No continuity either

Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com


Issues when graphing large networks1

Issues when Graphing Large Networks

Conclusion:

User Interaction

Querying

Selection and linking

Varying plot characteristics

--

Focus + Context

Pan & Zoom


References

References

  • Visualization 2003 - Network Visualization Course - T. Alan Keahey -www.visintuit.com

  • Graphics of Large Datasets: Visualizing a Million -- By Antony Unwin, Martin Theus and Heike Hofmann


  • Login