o p e n q u e stio n s in u n ce rta in ty v isu a liza tio ncscads.rice.edu/potter.pdf¥ exp e rime...
TRANSCRIPT
![Page 1: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/1.jpg)
Open Questions in Uncertainty Visualization
CScADS WorkshopKristin PotterJuly 29, 2010
![Page 2: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/2.jpg)
Advanced Computing and Scientific Data
• More bandwidth, storage, & computational power
• Larger data sets:- Higher resolutions- Longer runs
- More sophisticated models
2
Jaguar, ORNL
All this leads to huge amounts of complex data
![Page 3: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/3.jpg)
Uncertainty in Data• Scientific data sets are incomplete without
indications of uncertainty• Umbrella term for error, accuracy,
confidence level, missing data, inconsistencies, etc
• Multiple definitions depending on field or application
• Fundamental in science, why not in vis?
3
![Page 4: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/4.jpg)
Visualization is Communication• Translate data into images, “see” the data• Brings out relationships & features in data• Lets scientists communicate within their
fields and out to others
4
![Page 5: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/5.jpg)
Uncertainty Vis is Hard!• Adding more info to already large data
• Visual complexity and clutter• Can obscure data• Increasing visual “uncertainty” can
decrease understanding• What is an appropriate visual
metaphor?• No singular definition, no singular
solution5
![Page 6: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/6.jpg)
Understanding Uncertainty
• Influential in reasoning, decision making, and risk analysis
• Sources throughout the scientific process: acquisition, transformation, sampling, quantization, interpolation, classification, visualization...
• Provenance of uncertainty important in understanding
6
![Page 7: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/7.jpg)
Types of Uncertainty
• Experimental Uncertainty- NIST defines uncertainty
as standard deviation of a measurand*
• Geometric Uncertainty• Simulation Uncertainty• Visualization Uncertainty
7
* Barry N. Taylor and Chris E. Kuyatt. Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results. NIST Technical Note 1297, 1994.
![Page 8: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/8.jpg)
Types of Uncertainty
• Experimental Uncertainty• Geometric Uncertainty
- Unknowns in spatial positions
• Simulation Uncertainty• Visualization Uncertainty
8
* S. Lodha, B. Sheehan, A. Pang and C. Wittenbrink. Visualizing Geometric Uncertainty of Surface InterpolantsIn Proceedings of Graphics Interface '96, pp. 238--245. 1996.
![Page 9: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/9.jpg)
Types of Uncertainty
• Experimental Uncertainty• Geometric Uncertainty• Simulation Uncertainty
- Multimodel, ensembles or non-determininistic
• Visualization Uncertainty
9
* K. Potter, A. Wilson, P.T. Bremer, D. Williams, C. Doutriaux, V. Pascucci, C. JohnsonEnsemble-Vis: A Framework for the Statistical Visualization of Ensemble Data.In IEEE Workshop on Knowledge Discovery from Climate Data, pp. 233-240, 2009.
![Page 10: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/10.jpg)
Types of Uncertainty
• Experimental Uncertainty• Geometric Uncertainty• Simulation Uncertainty• Visualization Uncertainty
- Parameters of technique lead to differences
10
* C. Lundström, P. Ljung, A. Persson, and A. Ynnerman, Uncertainty Visualization in Medical Volume Rendering Using Probabilistic Animation,In IEEE TVCG, 13(6,) pp. 1648-1655, 2007,
![Page 11: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/11.jpg)
Types of Uncertainty
11
* C. Lundström, P. Ljung, A. Persson, and A. Ynnerman, Uncertainty Visualization in Medical Volume Rendering Using Probabilistic Animation,In IEEE TVCG, 13(6,) pp. 1648-1655, 2007,
• Experimental Uncertainty• Geometric Uncertainty• Simulation Uncertainty• Visualization Uncertainty
- Parameters of technique lead to differences
![Page 12: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/12.jpg)
Uncertainty Visualization
• Visually depict uncertainties• Faithfully present data• Improve vis as a decision making tool• Top visualization research problem *
12
* Chris R. Johnson. Top Scientific Visualization Research Problems,In IEEE CG&A 24(4) pp. 13--17, 2004.
![Page 13: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/13.jpg)
Approaching the Problem
• What is the nature of the uncertainty?• Is it a primary or secondary attribute?• Does the visualization design agree
with the data characteristics?
13
![Page 14: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/14.jpg)
Sensitivity-Type Analysis
• Input perturbations reflected in output
• Sensitivity of parameters
• Location & magnitude of variation important
14
![Page 15: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/15.jpg)
Multi-Model Ensemble Runs
• Collection of models predicting the same variable, time step, location
• Uncertainty in the variation of the models
• Standard deviation may not fully describe uncertainty
15
![Page 16: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/16.jpg)
Approaching the Problem
• What is the nature of the uncertainty?• Is it a primary or secondary attribute?• Does the visualization design agree
with the data characteristics?
16
![Page 17: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/17.jpg)
Primary Uncertainty
• Top-level information
• Where is the surface location?
• What is the boundary or range between tissue types?
17
![Page 18: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/18.jpg)
Primary Uncertainty
• Top-level information
• Where is the surface location?
• What is the boundary or range between tissue types?
18
![Page 19: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/19.jpg)
Secondary Uncertainty
• Annotation lines indicate missing data
• Minimal interference• No extra emphasis
on uncertain areas
19
* Andrej Cedilnik and Penny Rheingans.Procedural Annotation of Uncertain Information.In Proceedings of Vis ʼ00, pp. 77--84, 2000.
![Page 20: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/20.jpg)
Approaching the Problem
• What is the nature of the uncertainty?• Is it a primary or secondary attribute?• Does the visualization design agree
with the data characteristics?
20
![Page 21: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/21.jpg)
Uncertainty as a Scalar Value
21
* Gevorg Grigoryan and Penny Rheingans. Point-Based Probabilistic Surfaces to Show Surface UncertaintyIn IEEE TVCG, 10(5), pp. 546--573, 2004.
• Clear visual metaphor
• People can interpret blur and fuzz as uncertainty
• But they cannot quantify the amount of uncertainty from blur
![Page 22: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/22.jpg)
The Problem of Pretty Vis
• Reconstruction of medieval architecture• Shiny pictures, solid lines indicate truth• Sketchiness, opacity convey uncertainty
22
* Thomas Strothotte and Maic Masuch and Tobias Isenberg. Visualizing Knowledge about Virtual Reconstructions of Ancient Architecture. In Proceedings of Computer Graphics International, pp. 36--43, June, 1999.
![Page 23: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/23.jpg)
Solutions?
• Simplfication:• summarization, feature detection,
dimension reduction• Interaction:
• drill-downs, linked views, small multiples
• Flexibility:• use the right display for the right data
23
![Page 24: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/24.jpg)
Ensemble-Vis
• Multiple linked displays• user driven analysis
• Summary overviews• colormaps &
contours• Drill down
• 2D charts, direct data display
24
![Page 25: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/25.jpg)
What about evaluation?
• Missing for visualization in general• Typically user surveys, expert
assessment, anecdotal judgement • Influenced by personal preferences,
user experience, cultural biases, resistance to change
25
Better methods needed for ALL Vis!
![Page 26: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/26.jpg)
Take Home
• Qualitative information essential• Design should reflect sources, types, &
importance in application• Evaluation methods sorely needed
Each problem is unique & different: general approaches can only get you so far!
26
![Page 27: O p e n Q u e stio n s in U n ce rta in ty V isu a liza tio ncscads.rice.edu/Potter.pdf¥ Exp e rime n ta l U n ce rta in ty ¥ G e o me tric U n ce rta in ty ¥ Si mu la tio n U n](https://reader036.vdocuments.net/reader036/viewer/2022062611/6128c4eab60280415d4cd151/html5/thumbnails/27.jpg)
Thanks!
Questions?
27