Download - David S. Ebert ebertd@purdue
David S. [email protected]
Visual Analytics to Enable
Discovery and Decision Making:
Potential, Challenges, and Directions
Some material courtesy of Alan MacEachren, Bill Ribarsky, Antonio Sanfilippo, Kelly Gaither, Min Chen, Tom Ertl, Sonia Lasher-Trapp, Daniel Keim
September 2011
SFU, JIBCUBC
Ind U
Navajo Tech
UW
Stanford
GaTech
FIU
JSU
UT UHD Austin
U Stuttgart
VaTech
NC UNCCA&T
Penn St.
Swansea U
Purdue
September 2011
Motivation
To solve today’s and tomorrow’s problems requires exploring, analyzing, and reasoning with massive, multisource, multiscale, heterogeneous, streaming data
Image of Analyst’s Notebook
September 2011
Atmospheric Science: Multi-scale Interactions (in the words of a cloud physicist)
No observing platform can measure the quantities of interest over all needed spatial and temporal scales needed
No numerical model can simulate the quantities of interest over all needed spatial and temporal scales
We observe/simulate over a subset of the pertinent scales, using different instruments/models, and must assimilate these results to understand the “big picture”
Visual analytics is crucial for this task
Issues:Issues: Multi-scale, multi-system, multisource, massive, data & simulations
1 mm1 kHz
1km5min
September 2011
One Solution in Use: Our Atmospheric Visual Analytic EnvironmentUtilize multiple rendering styles
Provide interactive data exploration and user directed analysis
Allow user specified analysis and queries on the fly
Allow interactive correlative analysis of multisource data
September 2011
What Visual Analytics Enables
•Enable effective decision making through interactive visual analytic environments
•Enable effective communication of information
•Provide quantitative, reliable, reproducible evidence
•Enable user to be more effective from planning to detection to response to recovery
•Enable proactive and predictive visual analytics
September 2011
What’s Needed for Proactive and Predictive Visual Analytics?
•Reliable and reproducible models and simulation•Understanding of the data
• Distribution and skewness, errors, appropriate analysis techniques
•Understanding of the sources and types of data•Comparable or Correlative sources data
• Appropriate transformations applies to enable meaningful comparison and correlation
•Understanding of the use and problem to be solved!
September 2011
Four Challenges for Proactive & Predictive Visual Analytics at Scale
1. Computer-human visual cognition environments
2. Interactive simulation and analytics
3. Specific scale issues
4. Uncertainty and time
September 2011
Integrated Computer-Human Visual Cognition Environments
Balance of automated computerized analysis and human cognition to amplify human-centered decision making
Leverage both• Human knowledge and visual analysis to
increase analytical efficiency and guide simulations and analysis
• Interactive simulations, dimensional reduction, clustering, analytics to improve decision making
Create interactive discovery, planning & decision making environments
Discover knowledge about role of visual display and interfaces in discovery and decision-making
September 2011
Integrated Interactive Simulations and Analysis
Analysis and simulation must be interactive for integration into interactive environment
Need novel computational & statistical modelsGoal: enable improved discovery, decision making, analysis, and evaluation
September 2011
Visual Analytics At Real-World Scale
•Utilize advanced HPC techniques to enable interactive spatiotemporal analysis (spatiotemporal clustering, prediction)
• Cluster-based and cloud-based solutions
• GPGPU solutions
•Develop easily usable HPC visual analytic environments
•Example: Longhorn Exascale Visual Analytic Platform
• 2048 compute cores (Nehalem quad-core)
• 512 GPUs (128 NVIDIA Quadro Plex S4s, each containing 4 NVIDIA FX 5800s)
• 13.5 TB of distributed memory
• 210 TB global file system
September 2011
Scale: Multiscale Visual Analytics
Data at multiple semantic and physical scales must be integrated and analyzed to produce scalable solutions for all scales of the problem
Utilize natural problem scales
Enable cross-scale visual analysis
Enable decision making and action at all scales needed (e.g., neighborhood-city-state-nation, genome-cell-organ-body)
Interactive multisource, multiscale, multimedia analysis and integration of massive and streaming data
September 2011
Uncertainty and Temporal VA Challenges
Integrated, interactive temporal analytics
• Novel, interactive temporal analytical techniques
Intuitive reasoning and analysis across time and space
Precise information managing uncertaintyTemporal visual representations that provide context and do not introduce a propensity effect (e.g., from animation)
September 2011
Integrated Interactive Predictive Temporal Visual Analytics
Creating what-if and consequence evaluation environments with measures of certainty
Challenge:• Develop natural interactive visual spatiotemporal environments
–Seamless and natural interaction with and representation of temporal data
–Novel multivariate, multidimensional visual representations and analysis
September 2011
Result: Wise Visual Analytical Environments – Insight and Answers
Adapt analytics to integrate and perform with user-specified • Context• Constraints and boundaries
Incorporate analyst’s knowledgeIncorporate resources for planning, discovery, action
"Where is the wisdom we have lost in knowledge? Where is the knowledge we have lost in information?"
T. S. Eliot
September 2011
Result: Wise Integrated Interactive Predictive Visual Analytics
Challenges
• Scalable representation across problem scales
and user scales
• User-guided correlative and predictive analysis
• New temporal, spatiotemporal, precise,
multivariate, and streaming analytical techniques
September 2011
Keys for Success
•User and problem driven•Balance human cognition and automated analysis and modeling• Often applied on-the-fly for specific components identified by the user
•Interactivity and easy interaction • Utilizing HPC and novel analysis approaches
•Understandability of why predicted value is what it is
•Intuitive visual cognition•Not overloaded with features
September 2011
For Further Information
www.VisualAnalytics-CCI.org