Slycat Ensemble Analysis Patricia J. Crossno, Timothy M. Shead, Milosz A. Sielicki, Warren L. Hunt, Shawn Martin, and Ming-Yu Hsieh Sandia National Laboratories.

Slides:



Advertisements
Similar presentations
© Telelogic AB [1] Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company for the United States Department of Energys.
Advertisements

Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energys National Nuclear.
Conclusion Kenneth Moreland Sandia National Laboratories Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company,
Timothy M. Shead Sandia National Laboratories
Photos placed in horizontal position with even amount of white space between photos and header Sandia National Laboratories is a multi-program laboratory.
Introduction to Spreadsheets. Learning Target I can input data and do simple calculations in a spreadsheet.
Correlation and Linear Regression.
Ensemble Emulation Feb. 28 – Mar. 4, 2011 Keith Dalbey, PhD Sandia National Labs, Dept 1441 Optimization & Uncertainty Quantification Abani K. Patra, PhD.
Molecular Simulations of Metal-Organic Frameworks
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation,
Creating a Histogram using the Histogram Function.
1 Approved for unlimited release as SAND C Verification Practices for Code Development Teams Greg Weirs Computational Shock and Multiphysics.
Mapping Nominal Values to Numbers for Effective Visualization Presented by Matthew O. Ward Geraldine Rosario, Elke Rundensteiner, David Brown, Matthew.
Exploring Communication Options with Adaptive Mesh Refinement Courtenay T. Vaughan, and Richard F. Barrett Sandia National Laboratories SIAM Computational.
Visual Aids A quick and dirty primer Visual Aids Audience Advantages –Add clarity –Indicate what’s important –Reinforce key points –Increase interest.
Improving Contaminant Mixing Models For Water Distribution Pipe Networks Siri Sahib S. Khalsa University of Virginia Charlottesville, VA
3 CHAPTER Cost Behavior 3-1.
Dax: Rethinking Visualization Frameworks for Extreme-Scale Computing DOECGF 2011 April 28, 2011 Kenneth Moreland Sandia National Laboratories SAND P.
Differential Analysis & FDR Correction
Data Analysis Lab 04 Regression and Multiple Regression.
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation,
SciDAC SSS Quarterly Report Sandia Labs August 27, 2004 William McLendon Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed.
LAMMPS Users’ Workshop
Tables and Graphing Chapter 2 Section 3. Tables Tables- these display information in rows and columns so that it is easier to read and understand. Many.
Sandia is a multi-program laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear.
Site Report DOECGF April 26, 2011 W. Alan Scott Sandia National Laboratories Sandia National Laboratories is a multi-program laboratory managed and operated.
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation,
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation,
Multifidelity Optimization Using Asynchronous Parallel Pattern Search and Space Mapping Techniques Genetha Gray*, Joe Castro i, Patty Hough*, and Tony.
Graphs help us visualize numerical data.
DAY 6: EXCEL CHAPTERS 5 Rohit September 2 nd,
Scientific Notation. = 5.4 =3.47 Write the following in standard form A 1.8 X 10⁴ B 3.47 X 10⁷ C 4.3 X 10⁰ D 5.4 X 10⁻⁴ E 5 X 10⁻⁶ F (6 X 10⁴) (7 X 10⁵)
Correlation and Regression Stats. T-Test Recap T Test is used to compare two categories of data – Ex. Size of finch beaks on Baltra island vs. Isabela.
Jack Flicker, Robert Kaplar, Matt Marinella, and Jennifer Granata Sandia National Laboratories Acknowledgements Contact Sandia National Laboratories is.
CHART SELECTION TOOLKIT How To Choose The Right Chart For Your Audience.
SciDAC SSS Quarterly Report Sandia Labs January 25, 2005 William McLendon Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed.
A Kriging or Gaussian Process emulator has: an unadjusted mean (frequently a least squares fit: ), a correction / adjustment to the mean based on data,
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation,
APPLICATIONS FOR STRATEGIC ASSESSMENT,
EXPLORATORY DATA ANALYSIS and DESCRIPTIVE STATISTICS
Color Marking Body Response Lab
Click to edit Master title style
Low Voltage Ride Through (LVRT)
Color Marking Walking Lab
Craig Olson, Tina Tanaka, Tim Renk, Greg Rochau, Robert Peterson
Ray-Cast Rendering in VTK-m
A quick and dirty primer
MASTERING CHART SELECTION
Attention Narrows Position Tuning of Population Responses in V1
Results of Eddy Current Analysis
Hawaii Energy Storage Seminar: Other Energy Storage Technologies
DataLyzer® Spectrum SPC Wizard.
“Coding” for the building blocks of our bodies Edited
Average Number of Photons
Chemical Safety & Security Standard Operating Procedures
Align The Stars Continue.
Align The Stars Continue.
Align The Stars Continue.
Norm-Based Coding of Voice Identity in Human Auditory Cortex
Volume 90, Issue 6, Pages (June 2016)
Directions: Color your name with any tiles you want.
Γ-TEMPy: Simultaneous Fitting of Components in 3D-EM Maps of Their Assembly Using a Genetic Algorithm  Arun Prasad Pandurangan, Daven Vasishtan, Frank.
Align The Stars Continue.
Align The Stars Continue.
Patrick Kaifosh, Attila Losonczy  Neuron 
Align The Stars Continue.
Building and Linking against Trilinos
Γ-TEMPy: Simultaneous Fitting of Components in 3D-EM Maps of Their Assembly Using a Genetic Algorithm  Arun Prasad Pandurangan, Daven Vasishtan, Frank.
Lecture 16. Classification (II): Practical Considerations
Patrick Kaifosh, Attila Losonczy  Neuron 
Presentation transcript:

Slycat Ensemble Analysis Patricia J. Crossno, Timothy M. Shead, Milosz A. Sielicki, Warren L. Hunt, Shawn Martin, and Ming-Yu Hsieh Sandia National Laboratories Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL SAND P Patricia J. Crossno: Timothy M. Shead: Milosz A. Sielicki: Warren L. Hunt: Shawn Martin: Ming-Yu Hsieh: Analysis Tasks: Find strongest input/output correlations Find inputs with least impact on outputs Find anomalous simulation runs CCA Visual Representations Scatterplot: Each Simulation Relative to Ensemble Distance off diagonal shows difference from ensemble as a whole, plus potential anomalies. Purple = Outputs Bar chart: Ensemble-wide Relationships Viewing 1 st CCA component in both views Input x1 has the least impact on outputs y1 and y2 250 simulations, each color-coded by its y1 output value Selected simulation Positive many-to- many correlation (bar color the same) between X25 & X14 and Y2 & Y1 Green = Inputs Inputs x25 & x14 have the most impact on both outputs y1 and y2 Viewing 2 nd CCA component in both bar chart & scatterplot 250 simulations, each color-coded by its x23 input value Inputs and outputs sorted by correlation strength within CCA2 component X23 selected for scatterplot color- coding (dark green row highlight) Three distinct groups of input values Inverse correlation (red vs. blue) between x23 & y4; CCA3 captures relationship between x8 & y3 Scatterplot color- coding changed by clicking on y4 row (darker purple highlight) Three output value groups map to the 3 input groups 250 simulations, each color-coded by its y4 output value Click CCA column header to select CCA component in views Viewing 3 rd CCA component in both bar chart & scatterplot Inverse correlation between x8 & y3; CCA2 captures relationship between x23 & y4 250 simulations, each color-coded by its x8 input value X8 inputs range from low (blue) to high (red) X8 selected for scatterplot color- coding (dark green row highlight) Click header triangle to sort variables (toggles from decreasing to increasing) 250 simulations, each color-coded by its y3 output value Corresponding y3 outputs inversely range from high (red) to low (blue) Scatterplot color- coding changed by clicking on y3 row (darker purple highlight) Approach: Canonical Correlation Analysis (CCA) features simulations outputs inputs s1s1 s2s2 snsn o2o2 i1i1 omom … s3s3 s4s4 ikik o1o CCA features inputs i1i1 ikik outputs o2o2 o1o omom CCA components c1c1 ckck … CCA1 input meta- features output meta- features s1s1 s2s2 snsn s4s4 s3s3 Structure Correlations Slycat Sensitivity Analysis Input parameters SimulationEnsemble Simple Regression (1-to-1) Multiple Regression (Many-to-1) Model Confidence How About Many-to-Many Correlations? Problem: Electrical Circuit Simulation Sensitivity Analysis Rerun CCA analysis between all inputs and y4 to find strongest correlations (all-to-1) All to y4 analysis 4 anomalous runs share common x248 values 2641 simulations, each color- coded by its x248 input value (strongest) All to y4 analysis 4 anomalous runs share common x255 values 2641 simulations, each color-coded by its x255 input value (2 nd strongest) 2641 simulations, each color-coded by its y4 output value 4 anomalous runs in y4 values All to all analysis Finding Anomalous Simulations Finding Most Significant Inputs Objectives: Map Output Variability Back to Inputs Reduce Number of Input Parameters Reduce Number of Simulations to Run Identify Anomalous Runs Increase Model Confidence 266 scrollable Inputs Note R2 is increasing & P is decreasing with each CCA component Available Open Source Reduce Inputs & Simulations In the 2641 run ensemble above, analysis allowed input parameters to be reduced from 266 to 21, decreasing simulation time ten-fold.