lineup - sci.utah.edusci.utah.edu/~vdl/papers/2013_infovis_lineup_talk.pdf · 10.optimization....

Nils Gehlenborg Hanspeter PfisterAlexander LexMarc StreitSamuel Gratzl

LineUpVisual Analysis of Multi-Attribute Rankings

Rankings are omnipresent

2

Goal

Samuel Gratzl, Johannes Kepler University, Austria 3

IntuitiveInteractive

Multi-Attribute

Ranking Visualization

Goal

Samuel Gratzl, Johannes Kepler University, Austria 3

IntuitiveInteractive

Multi-Attribute

Ranking VisualizationTo Create

RefineExplore

10 Requirements

Samuel Gratzl, Johannes Kepler University, Austria 4

LineUp

Samuel Gratzl, Johannes Kepler University, Austria 5

LineUp

Samuel Gratzl, Johannes Kepler University, Austria 5

“It’s really complex to

make something simple…”Jack Dorsey (creator of Twitter)

Samuel Gratzl, Johannes Kepler University, Austria 6

Samuel Gratzl, Johannes Kepler University, Austria 7

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Samuel Gratzl, Johannes Kepler University, Austria 8

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

10 Requirements

Encode Rank

Samuel Gratzl, Johannes Kepler University, Austria 9

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

10 Requirements

1. Encode Rank

Samuel Gratzl, Johannes Kepler University, Austria 10

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Samuel Gratzl, Johannes Kepler University, Austria 11

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

10 Requirements

1. Encode Rank

Encode Cause of Rank

Samuel Gratzl, Johannes Kepler University, Austria 12

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

10 Requirements

1. Encode Rank2. Encode Cause of Rank

Samuel Gratzl, Johannes Kepler University, Austria 13

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score

84.2

44.0

64.3

73.8

89.4

Samuel Gratzl, Johannes Kepler University, Austria 13

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score

Samuel Gratzl, Johannes Kepler University, Austria 14

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

Support Multiple Attributes

Samuel Gratzl, Johannes Kepler University, Austria 15

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score

10 Requirements

1. Encode Rank

2. Encode Cause of Rank3. Support Multiple Attributes

Samuel Gratzl, Johannes Kepler University, Austria 16

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score A B C

Score = f(A, B, C)

Combiner functions: f(A,B,C)

(Weighted) sumScore = wa A + wb B + wc C

Samuel Gratzl, Johannes Kepler University, Austria 17

Combiner functions: f(A,B,C)

(Weighted) sumScore = wa A + wb B + wc C

MaximumScore = max(A, B, C)

Samuel Gratzl, Johannes Kepler University, Austria 17

Combiner functions: f(A,B,C)

(Weighted) sumScore = wa A + wb B + wc C

MaximumScore = max(A, B, C)

Product

Nesting

…

Samuel Gratzl, Johannes Kepler University, Austria 17

Combiner functions: f(A,B,C)

(Weighted) sumScore = wa A + wb B + wc C

MaximumScore = max(A, B, C)

Product

Nesting

…

Samuel Gratzl, Johannes Kepler University, Austria 17

Serial

Parallel

ComplexCombiners

Serial Combiner

Samuel Gratzl, Johannes Kepler University, Austria 18

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

A B C

wa A + wb B + wc C

Serial Combiner

Samuel Gratzl, Johannes Kepler University, Austria 18

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

A B C

wa A + wb B + wc C

(as Stacked Bar)

Serial Combiner

Samuel Gratzl, Johannes Kepler University, Austria 19

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

A B C

(as Stacked Bar)

wa A + +wb B wc C

Serial Combiner

Samuel Gratzl, Johannes Kepler University, Austria 19

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

A B C

(as Stacked Bar)

wa A + +wb B wc C

Serial Combiner

Samuel Gratzl, Johannes Kepler University, Austria 20

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

Princeton, USA

A B C

(as Stacked Bar)

wa A + +wb B wc C

Parallel Combiner

Samuel Gratzl, Johannes Kepler University, Austria 21

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

A B C

Parallel Combiner

Samuel Gratzl, Johannes Kepler University, Austria 21

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

AMAX(A,B,C)

(as Multi Bar)

Parallel Combiner

Samuel Gratzl, Johannes Kepler University, Austria 22

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

MAX(A,B,C)

(as Multi Bar)

MAX(A,B,C)

Parallel Combiner

Samuel Gratzl, Johannes Kepler University, Austria 23

UniversityRank

2.

5.

4.

3.

1. Oxford, UKHarvard, USA

MIT, USA

Cambridge, UK

Princeton, USA

(as Multi Bar)

MAX(A,B,C)

Parallel Combiner

Samuel Gratzl, Johannes Kepler University, Austria 25

UniversityRank

2.

5.

4.

3.

1. Oxford, UKHarvard, US

MIT, US

Cambridge, UK

Princeton, US

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple AttributesInteractive Refinementand Visual Feedback

MAX(A,B,C)

Parallel Combiner

Samuel Gratzl, Johannes Kepler University, Austria 25

UniversityRank

2.

5.

4.

3.

1. Oxford, UKHarvard, US

MIT, US

Cambridge, UK

Princeton, US

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinementand Visual Feedback

Samuel Gratzl, Johannes Kepler University, Austria 26

Samuel Gratzl, Johannes Kepler University, Austria 27

Video showing

• Creating combined score

• Sort by that

• Change weights

• Focus on animation: color & style

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinementand Visual Feedback

Flexible Mapping ofAttributes to Scores

Samuel Gratzl, Johannes Kepler University, Austria 28

Video showing

• Creating combined score

• Sort by that

• Change weights

• Focus on animation: color & style

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinementand Visual Feedback

5. Flexible Mapping ofAttributes to Scores

Samuel Gratzl, Johannes Kepler University, Austria 29

Min Max

0 1

Samuel Gratzl, Johannes Kepler University, Austria 29

0 100

0 1

Samuel Gratzl, Johannes Kepler University, Austria 30

0 100

0 1

Samuel Gratzl, Johannes Kepler University, Austria 310 100

01

Samuel Gratzl, Johannes Kepler University, Austria 32

Samuel Gratzl, Johannes Kepler University, Austria 33

Video showing

• Creating combined score

• Sort by that

• Change weights

• Focus on animation: color & style

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinementand Visual Feedback

5. Flexible Mapping ofAttributes to Scores

Compare Rankings

Samuel Gratzl, Johannes Kepler University, Austria 34

Video showing

• Creating combined score

• Sort by that

• Change weights

• Focus on animation: color & style

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support MultipleAttributes

4. Interactive Refinementand Visual Feedback

5. Flexible Mapping ofAttributes to Scores

6. Compare Rankings

Samuel Gratzl, Johannes Kepler University, Austria 35

University

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score

Samuel Gratzl, Johannes Kepler University, Austria 35

Rank

2.

5.

4.

3.

1.

ScoreUniversity

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score Score

Bump Charts

Samuel Gratzl, Johannes Kepler University, Austria 35

Rank

2.

5.

4.

3.

1.

ScoreUniversity

Harvard, USA

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

2.

5.

4.

3.

1.

Score Score

(+1)

(-2)

(+1)

Bump Charts

Samuel Gratzl, Johannes Kepler University, Austria 36

Rank

2.

5.

3.

1.

ScoreUniversity

Oxford, UK

Cambridge, UK

Princeton, USA

MIT, USA

Rank

5.

4.

3.

1.

Score Score

(+1)

(+1)

4.

Harvard, USA2.

(-2)

Samuel Gratzl, Johannes Kepler University, Austria 37

Samuel Gratzl, Johannes Kepler University, Austria 38

10 Requirements

6. Compare Rankings1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinementand Visual Feedback

5. Flexible Mapping ofAttributes to Scores

Scalability

Samuel Gratzl, Johannes Kepler University, Austria 39

10 Requirements

6. Compare Rankings1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinementand Visual Feedback

5. Flexible Mapping ofAttributes to Scores

7. Scalability

Scalability

Samuel Gratzl, Johannes Kepler University, Austria 40

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

Princeton, USA

A B C

Scalability

Samuel Gratzl, Johannes Kepler University, Austria 40

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

Princeton, USA

Score

Scalability

Samuel Gratzl, Johannes Kepler University, Austria 40

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

Princeton, USA

Scalability

Samuel Gratzl, Johannes Kepler University, Austria 40

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

Princeton, USA

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinement and Visual Feedback

5. Flexible Mapping of Attributes to Scores

Samuel Gratzl, Johannes Kepler University, Austria 41

6. Compare Rankings

7. Scalability

10 Requirements

1. Encode Rank

2. Encode Cause of Rank

3. Support Multiple Attributes

4. Interactive Refinement and Visual Feedback

5. Flexible Mapping of Attributes to Scores

Samuel Gratzl, Johannes Kepler University, Austria 41

6. Compare Rankings

7. Scalability8. Filtering9. Handle Missing Values10.Optimization

Optimization

Samuel Gratzl, Johannes Kepler University, Austria 42

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

A B C

Princeton, USA

Optimization

Samuel Gratzl, Johannes Kepler University, Austria 42

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

A B C

Princeton, USA

Optimization

Samuel Gratzl, Johannes Kepler University, Austria 42

University

Harvard, USAMIT, USA

Rank

2.

5.

4.

3.

1.

Oxford, UK

Cambridge, UK

A B C

Princeton, USA

Open source (BSD 3)

Source code on GitHub: https://github.com/Caleydo

Demos, videos, and more: http://lineup.caledyo.org

Implementation

Samuel Gratzl, Johannes Kepler University, Austria 43

http://caleydo.org

Summary

10 requirements for an intuitive interactive multi-attributeranking tool

LineUp visualization technique

Samuel Gratzl, Johannes Kepler University, Austria 44

Summary

10 requirements for an intuitive interactive multi-attributeranking tool

LineUp visualization technique

Future WorkWeb port

Integrate optimization algorithms

Samuel Gratzl, Johannes Kepler University, Austria 44

LineUp Applications

StratomeXCancer Subtype Analysis

45

EntouragePathway Analysis

LineUp Applications

StratomeXCancer Subtype Analysis

45

EntouragePathway Analysis

samuel.gratzl@jku.at

http://lineup.caleydo.org

LineUp

Samuel Gratzl, Alexander Lex, Nils Gehlenborg, Hanspeter Pfister, and Marc Streit

Visual Analysis of Multi-Attribute Rankings