Grinstein lecture/book visualizations to liven/explain theory paper
Jan 19, 2015
Grinstein lecture/book visualizations
to liven/explain theory paper
3.1. Preattentive Processing
Connection
2Source unknown
3.2. Theories of Preattentive Processing
3
Feature Integration Theory
(a) a boundary defined by a unique feature hue is preattentively classified as horizontal;
(b) a boundary defined by a conjunction of features cannot be preattentively classified as vertical
http://www.idvbook.com/
4Roland Rensink. “The Need for Attention to See Change.” http://www.psych.ubc.ca/∼rensink/flicker, March 2, 2003.
5Roland Rensink. “The Need for Attention to See Change.” http://www.psych.ubc.ca/∼rensink/flicker, March 2, 2003.
4. Perception in Visualization
6
Examples of perceptually motivated multidimensional visualizations:(a) visualization of intelligent agents competing in simulated e-commerce
auctions; (b) visualization of a CT scan of an abdominal aortic aneurism; (c) a painter-like visualization of weather conditions over the Rocky
Mountains
http://www.idvbook.com/
3.3. Feature Hierarchy
7
Example: Line Width
Source unknown
8
Bar Chart
Bar length better than area size (actually only area height was used!)
Solution from Stephen Few‘s Perceptual Edge
9
Improved Bar Chart
Solution from Stephen Few‘s Perceptual Edge
10
3.1. Position
Example visualizations: (left) using position to convey information. Displayed here is the minimum price versus the maximum price for cars with a 1993 model year. The spread of points appears to indicate a linear relationship between minimum and maximum price; (right) another visualization using a different set of variables. This figure compares minimum price with engine size for the 1993 cars data set. Unlike (left), there does not appear to be a strong relationship between these two variables.
http://www.idvbook.com/
11
3.2. Mark
This visualization uses shapes to distinguish between different car
types in a plot comparing highway MPG and
horsepower. Clusters are clearly visible, as well as
some outliers.http://www.idvbook.com/
12
3.3. Size (Length, Area and Volume)
This is a visualization of the 1993 car models data set, showing engine size versus fuel tank capacity.
Size is mapped to maximum price charged.
http://www.idvbook.com/
13
3.4. Brightness
Another visualization of the 1993 car models data set, this time illustrating the use of brightness to convey car width (the darker the points, the
wider the vehicle).
http://www.idvbook.com/
14
3.5. Color
A visualization of the 1993 car models, showing the use of color to display thecar’s length. Here length is also associated with the y-axis and is plotted against wheelbase. In this figure, blue indicates a shorter length, while yellow indicates a longer length.http://www.idvbook.com/
15
3.6. Orientation
Sample visualization of the 1993 car models data set depicting using highway milesper-gallon versus fuel tank capacity (position) with the additional data variable, midrange price, used to adjust mark orientation.
http://www.idvbook.com/
16
3.7. Texture
Example visualization using texture to provide additional information about the 1993 car models data set, showing the relationship between wheelbase versus horsepower (position) as related to car types, depicted by different textures.http://www.idvbook.com/
17
4.9. Senay and Ignatius (1994) VISTA
VISTA’s composition rules
Hikmet Senay and Eve Ignatius. “A Knowledge-Based System for Visualization Design.” IEEE Comput. Graph. Appl. 14:6 (1994), 36–47.
18
2. Two-Dimensional Data
A cityscape showing the density of air traffic over the United States at a particular time period.
LandscapesExample: News articles visualized as a landscape
• visualization of the data as perspective landscape
• the data needs to be transformed into a (possibly artificial) 2D spatial representation which preserves the characteristics of the data
High-Dimensional Data
Parallel Coordinates
Parallel Coordinates
Parallel Coordinates (Example)
BaseballLeague Database(1996)
Chernoff-FacesDo you spot any trend?
Stick Figures
5-dim. Imagedata from the
great lake region
25
4.3. Visualization Techniques
A storm cloud visualization containing glyphs showing wind direction and strength.
OpenDX (http://www.opendx.org/)
26
4.3. Visualization Techniques
Flow data visualized using ribbons, with vorticity mapped to twist.
OpenDX (http://www.opendx.org/)
27
4.3. Visualization Techniques
Corresponding points from several time slices can be joined to form streaklines.
OpenDX (http://www.opendx.org/)
28
1.1. Space-Filling Methods
A sample hierarchy and the corresponding treemap display.
Jing Yang, Matthew O.Ward, Elke A. Rundensteiner, and Anilkumar Patro. “InterRing: A Visual Interface for Navigating and Manipulating Hierarchies.” Information Visualization 2:1 (2003), 16–30.
29
1.1 Cushion Treemap
Idea: Use shading to construct a surface which shape encodes the tree structure.
The human visual system is trained to interpret variations in shade as illuminated surfaces .
see: H. van de Wetering and J. van Wijk. Cushion treemaps: Visualization of hierarchical information.In Proceedings of the IEEE Symposium on Information Visualization (InfoVis), 2005.
30
1.1 Newsmap
31
1.1 Treemap
Bederson, B.B., PhotoMesa: a zoomable image browser using quantum treemaps and bubblemaps, Proceedings of the 14th annual ACM symposium on User interface software and technology, pp 71-80, 2001, ACM
32
1.1. Space-Filling Methods
A sample hierarchy and the corresponding sunburst display.
Jing Yang, Matthew O.Ward, Elke A. Rundensteiner, and Anilkumar Patro. “InterRing: A Visual Interface for Navigating and Manipulating Hierarchies.” Information Visualization 2:1 (2003), 16–30.
33
2.1. Graphs-Drawing Conventions
Orthogonal
HierarchicalForce-Directed
Circular
Pictures from: www.tomsawyer.com
Edge oriented
Clustering oriented
Node oriented
Hierarchy oriented
34
Hierarchical Edge Bundling
35
Hierarchical Edge Bundling
More details in the paper:
• Bundling Strength• Alpha blending
Danny Holten, Hierarchical Edge Bundles: Visualization of Adjacency Relations in Hierarchical Data, IEEE TVCG, Vol 12, No 5, 2006 (Best Paper InfoVis 2006)
36
3.2. Tabular Displays
An example of Inxight Table Lens showing the cars data set sorted first by car origin and then by MPG.
Inxight Table Lens (http://www.inxightfedsys.com/products/sdks/tl/default.asp)
37
5.2. Hybrid Approaches
Example: XMDV Tool
XMDV allows to dynamically link and brush scatterplot matrices, star icons, parallel coordinates, and dimensional stacking (combination of geometric,
icon-based, hierarchical and dynamic techniques).
Matthew O. Ward, "Linking and Brushing.", Encyclopedia of Database Systems 2009: 1623-1626. http://davis.wpi.edu/xmdv/
38
5.2. Guidelines for Using Multiple Views
• Rule of Complementary:
Use multiple views when different views bring out correlations and/or disparities.
Georges Grinstein, UMass Lowell – Daniel Keim, Univ. of Konstanz – Matt Ward, WPI
42
1. Visualizing Spatial Data• Type of map depends on the properties of the
data, for example:
Dot maps Land use maps[2] Chloropleth maps
Line diagrams Isoline maps[3] Surface maps[1]
[1] K. Crane, Spin transformations of discrete surfaces, 2011[2] C. Power, Hierarchical fuzzy pattern matching for the regional comparison of land use maps , 2001[3] I. Solis, Isolines: energy-efficient mapping in sensor networks, 2005
8.3.1 Dot Map
A simple dot map of commercial wireless antennas in the USA43
44
2.1. Pixel Maps
The figures display U.S. Telephone Call Volume at four different times
during one day. The idea is to place the first data
items at their correct position and position
overlapping data points at nearby unoccupied
positions.
Daniel A. Keim, Christian Panse, and Mike Sips. “Visual Data Mining of Large Spatial Data Sets.” In Databases in Networked Information Systems, Lecture Notes in Computer Science, 2822, Lecture Notes in Computer Science, 2822, pp. 201–215.
Berlin: Springer, 2003.
0:00 am (EST) 6:00 am (EST)
10:00 pm (EST) 6:00 pm (EST) Overlap-free visualization!
45
3.2. Flow Maps and Edge Bundling
K.C. Cox. 3D geographic network displays. ACM Sigmod Record, 1996
Partially translucent arcs avoid overplotting.
The visualization of traffic flows of the United States to other countries suffers under visual clutter.
Arc maps try to avoid overlapping by mapping 2D lines into 3D arcs.
46
(a) Minard’s 1864 flow map of wine exports from France [20] (b) Tobler’s computer generated flow map of migration from California from 1995 - 2000. [18; 19] (c) A flow map produced by our system that shows the same migration data.
3.2. Flow Maps and Edge Bundling
D. Pahn et al. Flow map layout. Information Visualization, 2005.
Flow maps are used to show the movementof objects from one location to another.
They avoid overlapping by merging edges by, for example, clustering.
47
3.2. Flow Maps and Edge BundlingThe visualizations show IP flow traffic from external nodes on the outside to internal
nodes, visualized as treemaps on the inside. The edge bundling visualization (right side) significantly reduces the visual clutter compared to the straight line visualization (left
side).
Fabian Fischer, Florian Mansmann, Daniel A. Keim, Stephan Pietzko, and Marcel Waldvogel. “Large-Scale Network Monitoring for Visual Analysis of Attacks.” In Visualization for Computer Security: 5th International Workshop, VizSec 2008, Cambridge, MA, USA, September 15, 2008, Proceedings, Lecture Notes in Computer Science, 5210, pp. 111–118. Berlin: Springer- Verlag, 2008.
Flowstrates: Exploration of Temporal Origin-Destination Data
Ilya Boyandin, Enrico Bertini, Peter Bak, Denis Lalanne. Flowstrates: An Approach for Visual Exploration of Temporal Origin-Destination Data, EuroVis 2011
48
Applied “Force-Directed Edge Bundling”, Holten 200949
Making a visualization time-dependent
Every visualization can be made time dependent by providing several visualizations for several time points…
… in parallel … as a sequence (Animation)
10.2 Visualization techniques for serial data
1980 1990 2000
Time-Series Plot
One Parameter
Several Parameters
10.2 Visualization techniques for serial data
Gantt Chart10.2 Visualization techniques for serial data
LifeLines
LifeLines for medical records. Consultations, manifestations, documents, hospitalizations and treatments are shown in this record. Each doctor has a unique color. Line thickness shows severity and dosage.
10.2 Visualization techniques for serial data
History Flow
Editing history of the wikipedia „Microsoft“ page History flow visualization
Text of page
authors
10.2 Visualization techniques for serial data
ThemeRiverThemeRiver depicts thematic
variations over time within a large collection of documents
• directed flow from left to right movement through time
• horizontal distance between two points time interval
• total vertical distance collective strength of the selected themes
• colored currents individual themes
Data: Collection of patents from one company
10.2 Visualization techniques for serial data
Histogram vs. ThemeRiver
• Discrete values• Exact values• Hard to follow a single current
• Continuous flow• Interpolation, approximation• Easy to follow a single current (curving continuous lines)
10.2 Visualization techniques for serial data
Goal: Display large numbers of time series such that• relative importance and hierarchy relations can be quickly
perceived• the time series can easily be compared
(by arranging them in a regular layout)
Importance-Driven Visualization10.2 Visualization techniques for serial data
Importance-Driven Visualization
i-measure: volatility of stockscolor: normalized stock open price from green (low) through yellow (medium) to red (high)
80 time series from 9 different S&P500 Industries
10.2 Visualization techniques for serial data
Space-Time Cube
The space-time cube: I. An example of the author’s travels on an average Thursday in Enschede, the Netherlands. II. The space-time cube’s basics: a Space-Time Path and its footprint. The vertical line in the path represents the time a person remains at the same location, called station. III. A Space-Time-Prism (STP) indicates the locations that can be reached in a particular time interval (the Potential Path Space (PPS)). The projection of the PPS on the map results in the Potential Path Area (PPA).
10.2 Visualization techniques for serial data
Seesoft
color = statistic of interest, here: code age
Seesoft
Color is mapped to code age.
Three representations of code in the window: - text- line representation- pixel representation
ThemeScape Document Visualization
ThemeScape Document Visualization
A themescape representation of 700 articles related to the financial industry
Newsmap (Germany)
Text and Geo (1)
WS 2011 / 12 Computational Methods for Document Analysis, Prof. Dr. D. A. Keim 65
Chae et al. 2012
Seasonal Trend Decomposition
Word Clouds – http://wordle.net/
4 years of GK publications at the University of Konstanz(size of term corresponds to the frequency of the term within the publications)
Hyperbolic Browser
A hyperbolic browser representation of hierarchically ordered collection of documents
68
- techniques for selecting and highlighting objects and groups of objects
- often to identify the set of objects that will be the argument to some action
point is selected highlighted and can be dragged
1.2. Selection Operators
69
Dynamic Queries = visual means of specifying conjunctions
e.g.:
FilmFinderby C. Ahlberg and B. Shneiderman - sliders or radio buttons to select value ranges for variables in the Data Table
- cases for which all the variables fall into the specified ranges are displayed
1.3. Filtering Operators
70
1.3. Filtering Operators
Filtering rows and columns of the grades data set using XmdvTool.
XmdvTool (http://davis.wpi.edu/xmdv/)
71
cases that are selected in one view…
… are automatically also selected in all the other views
Screenshots of XMDV-Tool
interactive changes made in one visualization are automatically reflected in the other visualizations
1.6. Connection Operators
72
Overview & Detail
Overview
Detail
73
Documents arranged on a Perspective Wall
1. Screen SpacePerspective Wall
• The data outside the focal area are perspectively reduced in size
• The perspective wall is a variant of the bifocal lense display which horizontally compresses the sides of the workspace by direct scaling
74
original graph and fisheye view of the graph shows an area of interest quite large and with detail and the other areas successively smaller and in less detail graph visualization using a fisheye perspective
1. Screen Space - Fisheye
75
5. Data Structure Space
Example of shape simplification via dimension reordering. The left image shows the original order, while the right image shows the results of reordering to reduce concavities and
increase the percentage of symmetric shapes.
Wei Peng, Matthew O. Ward, and Elke A. Rundensteiner. “Clutter Reduction in Multi Dimensional Data Visualization Using Dimension Reordering.” In INFOVIS ’04: Proceedings of the IEEE Symposium on Information Visualization, pp. 89–96. Washington, DC: IEEE Computer Society, 2004.
76
TableLens with distortion (expansion) to show names
Visualization of a baseball database with a few rows being selected in full detail
6. Visualization Structure Space – TableLens
77
7. Animating Transformations
Example of a velocity curve corresponding to the position curve, with ease-in, ease-out
movement.
Example of an acceleration curve corresponding to the position curve, with
ease-in, ease-out movement.
3. System Performance - Use Case (1)
Practice Fusion Medical Research Data 15,000 de-identified health records, 7 different tables (patients, diagnosis, medications, etc.)
Data handling and visualization functionality evaluation
Task: visualize the distribution of women’s pregnancy age
79
3. System Performance - Use case (2)
VAST challenge 2011 1,023,057 geo-tagged microblogging messages with time stamps map information for the artificial “Vastopolis” metropolitan area
Geo-spatial-temporal data analysis functionality evaluation
Tableau Spotfire
Qlikview JMP
Task: visualize the geo-referenced disease outbreaks over the given time span
VAST 2013 Examples
Purdue UniversitySPRING RAIN
81
INRIA (Perin)
Arizona State University (Lu)
University of Konstanz (el Assady)
28
University of Konstanz (Schreck)
29
University of Stuttgart (Kruger)
30
Middlesex University (O’Connor-Read)
Central South University (Zhao)
Middlesex University (Choudhury)
92
Purdue University
93
SAS
Central South University
Peking University and Universität Stuttgart
University of Konstanz
University of North Carolina Charlotte