Enhanced Query-Driven Techniques for Uncertainty and Comparative Visualization

IIS-1018097

Principal Investigator: Kenneth I. Joy (kijoy@ucdavis.edu)
Co-Principal Investigator: Christoph Garth (cgarth@ucdavis.edu)

Award Period, 2010-2013, with extension through 2015

Last Updated: July 2015

 

Proposal Abstract:

Query-Driven Visualization (QDV) is a visualization-based discovery strategy that combines state-of-the-art methods from Scientific Data Management with modern visualization approaches to support rapid data analysis. By restricting computational and cognitive workload in visualization and interpretation to records defined to be significant by a scientist, fast visualization responses can answer intuitive questions about the data. Thus, query-driven techniques are ideal tools for data discovery and hypothesis testing. However, uncertainty in data and query can strongly and negatively influence a visualization result and hence the insight obtained from it. We provide a novel framework that generalizes QDV and is aimed at addressing deficiencies in existing methods. By modeling uncertainty at all levels of the query-driven process, we enable robust visualization techniques that incorporate the uncertain nature of the analysis process into the visualization result, thus enabling tools that will allow users to factor uncertainty into the conclusions drawn from data sets. The methods we develop are leverage multi-resolution data representations incorporating uncertainty information; this results in improved efficiency and parallel computation in answering queries over very large, high-dimensional data sets. We derive new visualization techniques by taking advantage of the improved flexibility, generality and efficiency of the provided framework, to address specifically the needs of comparative visualization of an ensemble of data sets pertaining to a common science problem. In order to increase the impact of our research, we will integrate our results with an open-source visualization package, distributing our work to a large community of scientists and engineers

Students/Collaborators on this Project:

Harald Obermaier, Postdoc, UC Davis, Now at Apple. Harald Obermaier is working on the uncertainty portion of the project, utilizing various statistical measures in visualization. He is lead author on two papers in the area in top conferences in the field during the past year. He has currently left the project and is employed by Apple.
Jennifer Chandler, UC Davis Jennifer Chandler, an NSF Fellowship winner, is working on the query system associated with this proposal. We expect to apply these uncertain queries to our ensemble models.
Kevin Bensema, UC Davis Kevin is working on initial development of ensemble visualization methods. He has been working collaboratively with partners at Pacific Northwest National Laboratory (where he is working in the summer months). He is also working with researchers in the climate area to use statistical methods for determining streams in the atmosphere.
Hank Childs, Lawrence Berkeley National Laboratory and the University of Oregon Hank was a professional research at UC Davis and the architect of VisIt, a visualization software system for scientific applications. He has taken a faculty position at the University of Oregon, and is a liaison with the Lawrence Berkeley National Laboratory's data analysis group who is collaborating with our project. We utilize VisIt for our technology transfer in this project.
Luke Gosink, Pacific Northwest National Laboratory We have been working with Luke Gosink at Pacific Northwest National Laboratory. Luke, as a graduate student at UC Davis, did some of the initial work on query-driven methods. He is currently providing data sets and problems on which we will base our research.
Christoph Garth, University of Kaiserslautern Christoph Garth is supporting this project through his work on uncertainty in flow. He is the co-PI of the project, and has taken a faculty position at the University of Kaiserslautern. Our efforts with Christoph have been very good over the past year, resulting in two joint publications.
Mathias Hummel, University of Kaiserslautern/ UC Davis Mathias is funded through the International Research Training Group of Kaiserslautern Germany, and visits the UC Davis Institute for Data Analysis and Visualization for approximately six months each year.
SDAV DoE Scalable Data Management, Analysis and Visualization (SDAV) Institute The SDAV Institute is a large-scale institute supported by the Department of Energy. We are collaborating with Lawrence Berkeley Laboratory scientists, working on ensemble portion of this proposal. Through this collaboration, we obtain relevant complex data sets and large-scale parallel systems to test our algorithms.
International Research Training Group, Kaiserslautern Germany Our Institute at UC Davis collaborates with the International Research Training Group in Kaiserslautern Germany. Typically three to five students visit us yearly and some work on projects related to this proposal. Mathias Hummel, who is currently at Davis, is a lead author on one of our papers under this project

Progress on the Project:

With this grant, we have addressed a difficult problem: Trying to apply statistical methods to visualization problems for large-scale data analysis. We have been working to integrate statistical queries into large-scale data visualization. The work has taken a slightly different direction by focusing on much more on the ensembles of data that scientists and engineers produce in large scale simulations. Here the scientists run many simulations with differing parameters, and need to use visual methods to analyze the outputs of these simulations. Many computational scientists believe that these "ensembles" of simulations will be the primary computational method to perform large-scale experiments in the future. We have found that the statistical methods that we have proposed in the grant, as well as this unique and relevant area of research, have given us many research problems to solve. Our recent papers develop new methods of visual analysis for problems in model generation uncertainty and in analysis ensembles that include flow. We have also used the methods developed from this work to develop new transfer function models for visualization and new methods for material interface generation.

The project has had substantial impact in the areas of ensembles as we have been setting up collaborations with researchers at three national laboratories to initiate new research efforts and to analyze data. We are pioneers in this area, and we believe that there is much fruitful work to do here.

The data sets produced by massive simulations are complex, and the mathematics/statistics required to solve them are also complex. We have made much progress to date.


Publications Submitted:

Kevin Bensema, Harald Obermaier and Kenneth I. Joy, "Visual Trends Analysis in Time-Varying Ensembles," Submitted to IEEE Transactions on Visualization and Computer Graphics.

Kevin Bensema, Luke Gosink, Harald Obermaier and Kenneth I. Joy, "Modality-driven Classification and Visualization of Ensemble Variance," submitted to IEEE Transactions on Visualization and Computer Graphics. (This paper been returned for a minor revision. We expect it to be accepted this summer.)

This paper utilizes a statistical test to determine modality in large scale ensemble data sets. It allows us to visualize the distributions developed at each data point, and determine their modality. This can tell the scientist if the distributions are acceptable for further analysis.


Publications to date:

Harald Obermaier and Kenneth I. Joy, "An Automated Approach for Slicing Plane Placement in Visual Data
Analysis," IEEE Transactions on Visualization and Computer Graphics, March 2015 issue.

Whereas this is not a paper related to ensembles, the methods we utilized were determined by our ensemble studies. The method allows one to use a slicing plane to visually scan a data set, and once fixed, it automatically adjusts the cutting plane to give the user more information about the data.

 

Mikhail M. Shashkov, Connie S. Hguyen, Mario Yepez, Mauricio Hess-Flores, and Kenneth I. Joy, "Semi-Autonomous Digitization of Real-World Environments," in Proceedings of the 19th International Conference on Computer Games: AI, Animation, Interactive Multimedia, Virtual Worlds and Serious Games (C-GAMES USA 2014) .

This paper has contributors from our undergraduate collaborators supported through our NSF supplements.

 

Harald Obermaier and Ronny Peikert, "Feature-Based Visualization of Multifields", Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization, Springer, 2014.

A survey of methods for visualizing complex data with uncertainty.

 

Harald Obermaier and Kenneth I. Joy. "Future Challenges for Ensemble Visualization", IEEE Computer Graphics and Applications, Vol. 34, No. 3, pp. 8-11, May/June 2014.

This paper appeared in Visualization Viewpoints in the IEEE Computer Graphics and Applications Journal, and discusses future research challenges for Ensemble Visualization.

 

Mathias Hummel, Harald Obermaier, Christoph Garth, and Kenneth I. Joy, "Comparative Visual Analysis of Lagrangian Transport in CFD Ensembles," IEEE Transactions on Visualization and Computer Graphics (Proceedings of IEEE Visualization 2013),  vol. 19, no. 12, pp. 2743-2752, 2013

This is the first paper that addressed the visualization of ensembles of time-varying flow fields. It won best scientific visualization paper at the IEEE Visualization Conference 2013.

 

Luke Gosink, Kevin Bensema, Hank Childs, Harald Obermaier and Kenneth I. Joy, "Characterizing and Visualizing Predictive Uncertainty in Numerical Ensembles Through Bayesian Model Averaging", IEEE Transactions on Visualization and Computer Graphics (Proceedings of IEEE Visualization 2013),vol. 19, no. 12, December 2013, 2703-2712.

Using Bayesian Model Averaging, we can compute a ``average'' solution to a problem, and compare queries against this average model. This solution is the first method that utilizes ground truth in the data to assist with the visualization.

 

Harald Obermaier and Kenneth I. Joy, "Local Data Models for Probabilistic Transfer Function Design in Interactive Volume Rendering", Proceedings of EuroVis Short Papers 2013 (to appear).

This paper utilizes Gaussian Mixture Models and Bayesian Methods to cluster points and, through an interactive process with the user, develop transfer functions for volume rendering applications

 

Iuri Prilepov, Harald Obermaier, Eduard Deines, Christoph Garth, and Kenneth I. Joy, "Cubic Gradient-Based Material Interfaces," submitted to IEEE Transactions on Visualization and Computer Graphics, vol. 19, no. 10, October 2013, 1687-1696.

This paper utilizes gradient surface methods to determine functions that generalize the "old" Youngs method of finding material interfaces for volume fraction data. The paper also describes methods by which we can make these surfaces continuous, within certain error constraints.

 

Simon Schroeder, John A. Peterson, Harald Obermaier, Louise H. Kellogg, Kenneth I. Joy and Hans Hagen, "Visualization of Flow Behavior in Earth Mantle Convection," IEEE Transactions on Visualization and Computer Graphics, vol. 18, no. 12, pp. 2198-2207, Oct. 2012

This paper was an application of statistical analysis to earth mantle data.

 

Luke Gosink, Christoph Garth, John Anderson, E. Wes Bethel, and Kenneth I. Joy, “An Application of Multivariate Statistical Analysis for Query-Driven Visualization,” IEEE Transactions on Visualization and Computer Graphics, Vol. 17, No. 3, 2011, 264-275.

Query-Driven Visualization (QDV) strategies are among the small subset of techniques that can address both large and highly complex datasets. This paper extends the utility of QDV strategies with a statistics-based framework that integrates non-parametric distribution estimation techniques with a new segmentation strategy to visually identify statistically significant trends and features within the solution space of a query.

 

Presentations:

Ken Joy was the keynote speaker at the Visualization and Data Analysis Symposium (VDA 2015). The titled of his talk was "Some Difficult Visualization Problems -- Big Science, Big Computer Systems, and Big Data"  
The Difficulties with Ensemble Visualization. A talk given by Ken Joy at the Dagstuhl 2014 Visualization Seminar, Dagstuhl Germany.  
Mathias Hummel presentation of the paper "Comparative Visual Analysis of Lagrangian Transport in CFD Ensembles." Given at the IEEE Visualization Conference 2013, Atlanta GA.  
Local Data Models for Probabilistic Transfer Function Design, by Harald Obermaier. Given at EuroVis 2013, Leipzig Germany  
Bayesian Methods Applied to Visualization Problems, by Ken Joy. Talk given at the Workshop on the Large-scale, High-Dimensional, and Multivariate Data using Topology and Statistics, June 2013, Le Barp France.  
Inverse Problems in Large-Scale Scientific Visualization, by Ken Joy. Talk given at the University of California, Davis, Statistical Sciences Symposium 2013: Analysis of Complex and Massive Data, April 2013  
Material Interface Reconstruction, by Ken Joy, Talk given at the University of Kaiserslautern Germany, Summer 2012  
Harald Obermaier's presentation of "Visualization of Material Interface Stability", given at the Pacific Visualization Conference, 2012.  
Harald Obermaier's presentation of "Function Field Analysis for the Visualization of Flow Similarity in Time-­‐Varying Vector Fields" at the International Symposium of Visualization Computing in Rethymnon, Crete, Greece. 2012  
The Impact of Data Complexity on Scientific Visualization Methods, Ken Joy, invited talk given at the SciDAC 2011 Conference.  
Integrated Multi-Field Visualization, by Ken Joy, given at the Dagstuhl Visualization Seminar, Leibnitz Center for Informatics, Dagstuhl Germany, June 2011  

Videos Related to the Project:

A video of spaghetti surfaces in the Pacific Northwest Laboratory Data Set, showing the difficulty of visualizing this data.  

Educational/Outreach Activities:

Mikhail M. Shashkov, Connie S. Hguyen, Mario Yepez, Mauricio Hess-Flores, and Kenneth I. Joy, "Semi-Autonomous Digitization of Real-World Environments," in Proceedings of the 19th International Conference on Computer Games: AI, Animation, Interactive Multimedia, Virtual Worlds and Serious Games (C-GAMES USA 2014) .

This paper has contributors from our undergraduate collaborators supported through our NSF supplements.

 
Ken Joy is giving a lecture in the Data Science Institute (a student summer program) at Los Alamos National Laboratory, July 23, 2014.  
The UC Davis Mobile App, initially developed by a team of students organized through the senior projects course has been transferred to the campus and is now the offical UC Davis app (both Android and IOS). This project initially started through student efforts on this proposal.  
Interactive Flythrough of the Known Universe. We sponsored this project through the UC Davis Computer Science Department's Senior Project Course. The students took a catalog of templates from the Sloan Digital Sky Survey, and wrote an octree-based sprite rendering system. They also utilize a LeapFrog system for navigation, and had 1.7M images rendering in real-time [60fps] off a MacBook Pro.  
Automatic image classification of the Miami skyline. We sponsored this project through the UC Davis Computer Science Department's Senior Project Course. A team of three students were able to develop a system that could process video from a camera system developed by Lawrence Berkeley Lab researchers and deployed in Miami. They read a bunch of literature from climate sciences and computer vision, and put together matlab code that automatically segment out, and classify cloud types with 95% accuracy. We expect to submit a poster to AGU this fall, with the students as co-authors.  

Online Course on Geometric Modeling, designed by Ken Joy, offered Fall 2012 at UC Davis.

This course has been expanded into a set of geometric modeling lectures with the purpose that any student worldwide can access these lectures and learn the basic material behind geometric modeling. Find the enhanced lectures here.

 
Images from the UC Davis Visualization and Graphics Art Show  
Why go to Graduate School Presentation, by Ken Joy -- given various times, UC Davis  
How to Give a Professional Talk -- given several times -- UC Davis.  

Awards:

Connie Nguyen, an undergraduate at UC Davis working on the supplementary funds of this grant, has won the Citation as the outstanding student in the Computer Science Department at UC Davis for 2014-15. She also has won a Scolar Award through the UC Davis College of Engineering for 2015.
Ken Joy won the IEEE TCVG Visualization Career award in October of 2014. This award was given in recognition of Ken's foundational research in the mathematical representation of data for visualization and for service to the community. The announcement of the award can be found here.
Harald Obermaier has been awarded an Honorable Mention in the competition for the UC Davis Award for Excellence in Postdoctoral Research. There are over 1000 postdocs at UC Davis eligible for this award.

The paper "Comparative Visual Analysis of Lagrangian Transport in CFD Ensembles," was selected Best Scientific Visualization paper of the IEEE Visualization Conference in 2013.

Kevin Bensema has been awarded a Graduate Student Research Fellowship by Lawrence Berkeley National Laboratory which covers the academic years 2013-2015. He has been awarded a summer internship from Pacific Northwest Laboratory for the summer of 2014.
Hank Childs has accepted a faculty position at the University of Oregon. He has recently received a positive tenure evaluation at the University.
Jennifer Chandler was awarded a 2012 NSF Graduate Fellowship
Christoph Garth has accepted a position of Junior Professor for Computational Topology at the University of Kaiserslautern in Germany, more information can be found at http://vis.uni-kl.de/people/garth/
Christoph Garth, co-PI of this project, was awarded the UC Davis Award for Excellence in Postdoctoral Research. This award is given yearly to two recipients, drawn from the nearly 1000 post docs at UC Davis.
The paper by Luke Gosink, Christoph Garth, John Anderson, E. Wes Bethel, and Kenneth I. Joy, entitled “An Application of Multivariate Statistical Analysis for Query-Driven Visualization,” and published in IEEE Transactions on Visualization and Computer Graphics, Vol. 17, No. 3, 2011, was selected as the spotlight paper for this journal for March 2011