View
0
Download
0
Category
Preview:
Citation preview
1
Beyond BallBeyond Ball --andand --StickStickPart 1: Using vision to thinkPart 1: Using vision to think
Mario ValleMario ValleSwiss National Supercomputing Centre (CSCS)Swiss National Supercomputing Centre (CSCS)
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Kekulé dreamKekulé dream
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Everyday mental modelsEveryday mental models
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Mental modelsMental models
“All our ideas and concepts are only internal pictures”
Ludwig Boltzmann (1899)
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Models for mental simulationModels for mental simulation
For example we use mental simulation (manipulation ofa mental model) to solve the problem:
“Which rotated image corresponds to the first one?”
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Models guide our perceptionModels guide our perception
The “cocktail” effect
Who called me?
2
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Models simplify and abstractModels simplify and abstract
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Where mental models liveWhere mental models live
Then the new concepts discovered crystallize as new knowledge building links to existing information
(Wickens memory and perception model)
Mental models are built in our working memory
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
ImaginationImagination
“Imagination is
more important
than knowledge”
Albert Einstein
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Imagination from where?Imagination from where?
“Imagination is vision running backwards”
S. Greenfield
Leo Leoni, Fish is Fish, Pantheon, 1970
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Working memory limitsWorking memory limits
Then the new concepts discovered crystallize as new knowledge building links to existing information
Mental models are built in our working memory
But working memory:
� Has limited capacity(7 ± 2 items)
� Disappears in ~2 min
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
External cognitionExternal cognition
We have always created extensions to our mind and body to overcome our limitations
3
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Example from mental mathExample from mental math
34 x72
------68
238-------2448
Time needed
0
20
40
60
80
100
120
Mental On paper
sec
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Vision and cognitionVision and cognition
At high level vision and cognition are linked.
We say “I see!” to mean “I understand!”
The visual system is an extension of our brain: 1/3 – 1/4 is dedicated to visual perception.
Therefore why don’t we use our visual system to comprehend numerical data?
Old IBM advertising
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Visual patterns discoveryVisual patterns discovery
Trends, Clusters, Gaps, Outliers, Correlations, ...
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Vision enables direct perceptionVision enables direct perception
The Mandelbrot set has a symmetrical structure that looks like an insect. Around the central body are placed various smaller scale replicas of the same set. The biggest replica is located on the left side of the main body. All around there are detail rich threadlike structures…
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Vision provides an holistic viewVision provides an holistic view
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
How many ‘3’ are there?How many ‘3’ are there?
89739057092794057962976509829408028085080830802809850-802808567847298872ty458202094757720021789843890r455790456099272188897594797902855892594573979209
4
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Immediately seen if color addedImmediately seen if color added
89739057092794057962976509829408028085080830802809850-802808567847298872ty458202094757720021789843890r455790456099272188897594797902855892594573979209
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Preattentive perceptionPreattentive perception
"Civilization advances by extending the number of important operations which we can perform without thinking about them."
Alfred North Whitehead
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Structure formationStructure formation
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
See nonexistent structuresSee nonexistent structures
After Garcia-M
ata & Shaffner (1934)
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Change blindness Change blindness
Our eye is not a movie camera!
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
““ Using vision to think”Using vision to think”
S. K. Card,J. D. Mackinlay,B. Shneiderman
5
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Few directly perceived phenomena Few directly perceived phenomena
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Most data remains not accessibleMost data remains not accessible
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Scientific visualization birthScientific visualization birth
“Visualization offers a method for seeing the unseen . It enriches the process of scientific discovery and fosters profound and unexpected insights. In many fields it is already revolutionizing the way scientists do science”
Visualization in Scientific Computing,McCormick et al.ACM SIGGRAPH, 1987
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Visualization definedVisualization defined
“[Visualization is] the use of computer-supported, interactive, visual representations of data to amplify cognition...”
Card, Mackinlay, and Shneiderman
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Uses color, shape, interactionUses color, shape, interaction
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Uses spatial metaphorsUses spatial metaphors
6
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
(too much) visual metaphors(too much) visual metaphors
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Natural hierarchy of newsNatural hierarchy of news
www.marumushi.com/apps/newsmap
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
The scientific discovery processThe scientific discovery process
EXPERIMENT ORDATA COLLECTION
DATA
Computation orTransformation
Rendered imageInsight
Hypothesis
User
interaction
guesswork
By J. Watson
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Visualization processVisualization process
Conceptualmodel
Conceptualmodel
DataStudy object
Datamodel
Datamodel
acquisition
Preconceptions &interpretation
Influen
ce
Interaction
Algorithms
RenderPerceptio
n
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Three visualization rolesThree visualization roles
1. Confirmatory visualization� Verify that some hypothesis holds
2. Exploratory visualization� Exploration-driven research
3. Presentation and communication� Present what has been discovered
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Classical data analysisClassical data analysis
problem
data
model
analysis
conclusions
• Focus on the model
• Hypothesis-driven research
• Quantitative methods
7
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Exploratory data analysisExploratory data analysis
problem
data
analysis
model
conclusions
• Focus on the data
• Exploration-driven research
• Graphical methods
• Evolutionary
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Present and communicate resultsPresent and communicate results
Francesco Gervasio –
ETH Zürich
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
The visualization mottoThe visualization motto
“Discover the unexpected, describe and explain the expected”
National Visualization and Analytics Center™Pacific Northwest National Laboratory
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Confused with Computer GraphicConfused with Computer Graphic
“But I was convinced that visualization is about creating nice images only!”
Goal of visualization is to improve cognition using visual methods, not to create illusion
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Limit visualization usefulnessLimit visualization usefulness
ModelingComputation
Visualization
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
An active role for visualizationAn active role for visualization
ModelingComputation
Visualization
8
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Visualization is a toolVisualization is a tool
EXPERIMENT ORDATA COLLECTION
DATA
Computation orTransformation
Rendered imageInsight
Hypothesis
User
interaction
guesswork
VISUALIZATION
VISUALIZATION
By J. Watson
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Ultimate goal: comprehensionUltimate goal: comprehension
Purpose of computing is insight, not numbers
Richard HammingNumerical Methods for Scientists and Engineers 1962
Purpose of visualization is insight, not pretty pictures
Visualization community
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
RecapRecap
� Visualization is a tool, an internal interface in the scientific discovery loop
� Visualization helps exploration besides presentation and hypothesis-driven research
� Visualization goal is not (only) to produce nice images, but to gain insight
Chemistry VisualizationChemistry Visualization
Now apply visualization principles to chemistry dat aNow apply visualization principles to chemistry dat a
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Chemistry segmentationChemistry segmentation
Materialscience
Crystallo-graphy
Genomic
Teachingsupport
Nanostructures
…
Moleculardynamics
ChemistryChemistry
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Chemists and modelsChemists and models
9
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Dogma: visualization helps insight Dogma: visualization helps insight
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Different clarity, why?Different clarity, why?
HEMOGLOBIN (VAL BETA1 MET, TRP BETA37 TYR) MUTANT
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Chemistry dataChemistry data
Conceptualmodel
Conceptualmodel
DataStudy object
Datamodel
Datamodel
acquisition
Preconceptions &interpretation
Influen
ce
Interaction
Algorithms
RenderPerceptio
n
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Do you remember GIGO?Do you remember GIGO?
GIGO /gi:'goh/ [acronym]1. Garbage In, Garbage Out : usually said in response to users who complain that a program didn't "do the right thing" when given imperfect input or otherwise mistreated in some way.
2. Garbage In, Gospel Out : this more recent expansion is a sardonic comment on the tendency human beings have to put excessive trust in “computerized” data.
Source: Jargon File 4.2.0
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Data influences visualizationData influences visualization
If you don’t know what the data represent this visualization is as good as any other.
Instead if you know the dataset content the visualization can foster insight because it is tuned to what the data represent.
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Chemistry data kindsChemistry data kinds
O
O
O
OH
Data from prof. A. Oganov – ETH Zürich
Structures
Scalar volumes
Spectra
Tables
10
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Time dependent dataTime dependent data
Images from AmiraMol
Sergey Churakov – PSI Villigen
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Death Receptor Signaling pathway
Gene expression
Not only numerical dataNot only numerical data
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Information VisualizationInformation Visualization
Scientific VisualizationPrimarily focused on physical dataData normally geometry basedScalar, vector and tensor data types
Information visualizationFocused on abstract and non physical dataNormally abstract, non geometric dataMultidimensional and non numeric data types
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Scientific visualization (streamlines)Scientific visualization (streamlines)
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Non spatial information Non spatial information
Not all data has an associated geometry (web site accesses, marketing data, etc.). Other have only an associated topology (trees, graphs).
A geometry should be assigned to visualize them.
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Think multidimensionalThink multidimensional
7 atoms x 3 coordinates each = trajectory in a 21-dimensional space
x1
x2
x3
x4
x21
11
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Represent dataRepresent data
Conceptualmodel
Conceptualmodel
DataStudy object
Datamodel
Datamodel
acquisition
Preconceptions &interpretation
Influen
ce
Interaction
Algorithms
RenderPerceptio
n
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Traditional representationsTraditional representations
� Ball & Stick (and its derivatives)
� Surfaces
� Spectra and line charts
� 2D charts & contours
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Simplified representationsSimplified representations
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Representation constraintsRepresentation constraints
“A stagnant set of representations limits the
way scientists think about their models and
thereby limits potential insights”
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Example: join the nine dots gameExample: join the nine dots game
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Wrong solution (5 lines)Wrong solution (5 lines)
12
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Right solution (4 lines)Right solution (4 lines)
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Unstated constraintUnstated constraint
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Why limit yourself?Why limit yourself?
http://www.gihanperera.com/mindgames/dots9ans.html
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Example: ice meltingExample: ice melting
Davide Donadio – ETH Zürich
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Example: mission impossibleExample: mission impossible
Unusual carbocations
More examples at: http://chemgroups.ucdavis.edu/~tantillo
Unbuildable molecules
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Visualization algorithmsVisualization algorithms
Conceptualmodel
Conceptualmodel
DataStudy object
Datamodel
Datamodel
acquisition
Preconceptions &interpretation
Influen
ce
Interaction
Algorithms
RenderPerceptio
n
13
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Visualization algorithmsVisualization algorithms
Data from prof. A. Oganov – ETH Zürich
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
3D is not everything3D is not everything
3D structure displaysare valuable tools, but ...
� limited to viewing part of structure
� unsuited for quick comparisons
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
2D plots are still important2D plots are still important
2D structure plots are still the coreof chemical information:
� show complete structure
� easy recognition of patterns
Chemists know how good structures looks like.
N
Ni
N
NN
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Try to reconstruct 3D from 2DTry to reconstruct 3D from 2D
Top view
Front view
Side view
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
3D to understand shape3D to understand shape
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
How to convey depth perceptionHow to convey depth perception
Without With depth cueing
14
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Shadows and modelsShadows and models
No shadow Hard shadow Soft shadowBeyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Using artificial depth cuesUsing artificial depth cues
Drop lines / projections� Can clutter the image
Occlusion� More natural� But hides part of the data
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Human perceptionHuman perception
Conceptualmodel
Conceptualmodel
DataStudy object
Datamodel
Datamodel
acquisition
Preconceptions &interpretation
Influen
ce
Interaction
Algorithms
RenderPerceptio
n
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Influence of color perception Influence of color perception Gaussian cube
with default colorm
ap
Adjusted data range
Better color mapping
Perceptually tuned
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Eye attractorsEye attractors
Smooth bonds Two colors bonds
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Interact with the representationInteract with the representation
Conceptualmodel
Conceptualmodel
DataStudy object
Datamodel
Datamodel
acquisition
Preconceptions &interpretation
Influen
ce
Interaction
Algorithms
RenderPerceptio
n
15
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Reduces 3D ambiguitiesReduces 3D ambiguities
Illusion due to thefixed viewpoint
The famous Ames Room optical illusion
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Lets interact and exploreLets interact and explore
Interactive element tableChanges to the atomic properties with the cursors are immediately reflected as highlighted elements
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Building a volume mental modelBuilding a volume mental model
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Use touch (and maybe smell…)Use touch (and maybe smell…)
Tangible model
The system adds to the real image other molecules or show the electrical field around the solid model http://www.scripps.edu/mb/olson/pyartk/pyartk.html
Breaking barriersBreaking barriers
Where the true power of visualization liesWhere the true power of visualization lies
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Data fusionData fusion
Data from different sources and of different kind can be fused together with great advantage:
� Create an interpretative context for the data
� Suggest correlations
� Highlight cause-effect relationships
16
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Charts
Data are not ‘single’Data are not ‘single’
Loosely correlated charts
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
2D GIS view
Charts
Data are not ‘single’Data are not ‘single’
Now fused in a 2D view plus GIS context map
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
2D GIS view
3D Data FusionCharts
Data are not ‘single’Data are not ‘single’
Complete (and useful) data
fused together
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Mix standard techniquesMix standard techniques
Electronic ring currents in benzene
Data simulation by Daniel Sebastiani, Max Planck Institute
Winds over Europe
Daily weather forecast model computed at CSCS for MeteoSwiss
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Mix standard techniquesMix standard techniques
Volume rendering
Vector glyphs, Line Integral Convolution (LIC) and alpha blending to cut the hole
Sergey Churakov – PSI Villigen
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Enlarge visualization spaceEnlarge visualization space
Continuum step
Atomistic step
17
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Example: fluid dynamicsExample: fluid dynamics
Vladimir Slavin – Brown University Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Consider other data typesConsider other data types
Brain diffusion tensors
Isotropic Component of NMR 31P Shielding Tensor
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Multidimensional visualizationMultidimensional visualization
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Multiple linked viewsMultiple linked views
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Data MiningData Mining
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
Crystallographic data miningCrystallographic data mining
Usage of computed simplified X-Ray structure factors to detect structural changes
Data from prof. A. Oganov – ETH Zürich
18
Beyond Ball-and-Stick Tutorial – Mario Valle – CSCS User Day 27/09/2005
RecapRecap
� The importance of choosing the right representation
� There are (virtual) limits from the usual chemistry visualization methods
� Big advantages from going beyond the usual horizons
Beyond BallBeyond Ball --andand --StickStick
Thanks for your attention!Thanks for your attention!
Mario ValleMario Valle
mvalle@cscs.chmvalle@cscs.chhttp://www.cscs.ch/~mvalle/ http://www.cscs.ch/~mvalle/
Recommended