Limn: Using Movie Technology to Drive Graphics for Large Data Manuel Suarez†, Dianne Cook†, Peter Sutherland‡, Justin McIllece†. † Iowa State University ‡ Affymetrix, Inc. Project Description Explore for multivariate structure in massive data. Limn (lîm), v.t.; to present an image or lifelike imitation of; Limit-n; to be able to explore any data set to its limits. 2 JSM 2001 - Atlanta August 8, 2001 Topics of Discussion 3 Visualizing Multivariate Data Extensions to Large Data Application Software Demo Limitations/Continuing Work JSM 2001 - Atlanta August 8, 2001 What is Multivariate Data? Matrix of data: cases=rows, variables=columns x11 x 21 X xn1 4 x12 x22 xn 2 JSM 2001 - Atlanta x1 p x2 p xnp nxp August 8, 2001 Visualizing Multivariate Data Multiple Views Paradigm, Augmented by: •Focusing using zoom/pan/re-scale •Linking by queries, or motion •Rearranging to make multiple comparisons 5 JSM 2001 - Atlanta August 8, 2001 Touring Algorithm Interpolate between a series of projection planes joint distribution/multivariate shape X ( nxp ) P( px 2 ) Y( nx 2 ) 6 JSM 2001 - Atlanta August 8, 2001 Touring Algorithm (cont.) 7 Grand: random choice of base planes Guided: functional choice of base planes Manual: User controlled interpolation Neil Sloane’s “Equi-distant” Bases JSM 2001 - Atlanta August 8, 2001 Visualization of Large Data Sets Data Reduction – – Scaling of Methods – – 8 Binning – Lose intricate structure Subset – Lose rare structure Computation Storage JSM 2001 - Atlanta August 8, 2001 Limn Software Two Steps: 1. 2. Create and animation sequence, and save as QuickTime movie or series of JPEG images View animation, and interact with it by brushing, and overlaying subsets of data. * Code is written in Java 9 JSM 2001 - Atlanta August 8, 2001 Data of Interest Multivariate Spatio-Temporal Data, e.g. continuously-recording monitoring stations or remotely sensed data. Movie Space S M T W Th F S Time 10 JSM 2001 - Atlanta August 8, 2001 Test Data Set 11 Monitoring global climate using Pacific Ocean moorings measuring temperature, wind, … JSM 2001 - Atlanta August 8, 2001 Demo Background Movie: All data, 1980-1998 Subset Overlays – – – – – 12 December 1993, Eastern Pacific (Normal year) December 1993, Western Pacific December 1997, Eastern Pacific (El Niño year) December 1997, Western Pacific Binned Data JSM 2001 - Atlanta August 8, 2001 Limitations/Continuing Work User Interface – – Create subsets by space and time Composite movie layers for subsets Database – – – 13 For large subsets For viewing density Distributed data and live updating of data Fast Indexing for linked views (e.g., space-time) Catalog movies of full data and subsets JSM 2001 - Atlanta August 8, 2001 Summary 14 A new approach to visualizing large, multivariate data Combine pre-computed tour movies of large data with real-time computed subsets Extends tour algorithm to arbitrarily large data sets JSM 2001 - Atlanta August 8, 2001 Contact Information www.public.iastate.edu/~dicook/Limn/index.html Email: dicook@iastate.edu suarezm@vrac.iastate.edu 15 JSM 2001 - Atlanta August 8, 2001