ay/bi199: methods of computational science visualization jumpstart+tools+techniques santiago v lombeyda center for advanced computing research caltech

Transcription

1 ay/bi199: methods of computational science visualization jumpstart+tools+techniques santiago v lombeyda center for advanced computing research caltech

2

3 analytical answers new questions data questions visual inspection answers

4 what includes visualization data more understandable representation

5 visualization = science + computer graphics/hci + graphic design/art

6 finding solution(s) via purpose for what purpose do we visualize? quick view/demonstration we want to look at/show something particular analysis we know what we are looking for explore we do not know what we are looking for debugging we want to assure there is nothing odd

7 process dictated by the data A CLOSER LOOK AT THE DATA

8 abstract data: GEOMETRIC STRUCTURE multi-dimensional data RECORDS mapping + paradigms interaction infoviz 2d/3d data scalar/vector/tensor + time paradigms interaction the more main stream viz ieee vis ieee infovis siggraph ieee xplore: ieeexplore.ieee.org acm digital library: portal.acm.org/dl.cfm

9 abstract 2d/3d data data: GEOMETRIC STRUCTURE MPG Cylinders Horsepower Weight Acceleration Year Origin

10

11 understanding V THE VISUALIZATION PROCESS

12 [ Ben Fry, Visualizing Data, O Reilly Media, 2008 ]

13 Acquire Parse Filter Mine Represent Refine Interact Obtain the data, whether from a file on a disk or a source over a network. Provide some structure for the data's meaning, and order it into categories Remove all but the data of interest. Apply methods from statistics or data mining as a way to discern patterns or place data in mathematical context. Choose a basic visual model, such as a bar graph, list, or tree. Improve the basic representa tion to make it clearer and more visually engaging. Add methods for manipulati ng the data or controlling what features are visible. [ Ben Fry, Visualizing Data, O Reilly Media, 2008 ]

14 usual visualization engine data processing algorithms geometry generation rendering ui interaction

15 the visualization toolkit VTK c/c++ tcl/tk data processing algorithms geometry generation rendering ui python java interaction R

16 interactive renderers OpenGL HIGHER LEVEL: mesagl directx OpenInventor C++/GL data processing algorithms geometry generation rendering ui COIN Java3D interaction Java/GL 03D (by Google) WebGL

17 ray tracers POVRay RenderMan $$ data processing algorithms geometry generation rendering ui MODELLERS: Blender interaction Maya $$

18 BLENDER

19 POVRAY

20 RENDERMAN

21 gui toolkits QT python GTK+ data processing algorithms geometry generation rendering ui python TK (TCL/TK) interaction Java Swing, Cocoa, Motiff

22 web based ui data processing algorithms geometry generation rendering ui dhtml (HTML5 or SVG)/ JavaScript Processing Java JProcessing interaction Flash

23 visualization system Paraview VTK LLNL VisIt VTK EnSight $ Protovis WWW data processing algorithms geometry generation rendering ui Many Eyes WWW Wiki based interaction Modrian R, TopCat, Mollegro,

24

25 an overview: tools & techniques INFOVIZ

26 abstract 2d/3d data data: GEOMETRIC STRUCTURE MPG Cylinders Horsepower Weight Acceleration Year Origin

27 basic infovis techniques: N-DIMENSIONAL RECORD DATA

28 visual analytics goal: (mapping data) detect, classify, and measure X trends, outliers, patterns, clusters, correlations

29 principal component analysis: pca = data transformation n correlated variables n uncorrelated variables (akin decomposition into orthogonal element basis)

30 choose a layout strategy..

31 PARALLEL COORDINATES + CLUSTERING

32 SCATTERPLOT MATRIX STAR PLOT GLYPHS

33 DROSOPHILIA 2D & 3D SCATTERPLOTS

34 DROSOPHILIA PARALLEL COORDINATES

35 3D PARALLEL COORDINATES

36 basic infovis techniques: HEIRARCHICAL DATA

37 TREEMAPS

38 HEIRARCHICAL RINGS

39

40 DPOSS & Fisher s Iris : MONDRIAN+R quick demo

41 infovis packages mondrian (R based) rosuda.org/mondrian xmdv davis.wpi.edu/~xmdv molegro data modeller (bio) topcat (astro) many eyes (online wiki) protovis/polaris (+DATACUBES) COMERCIAL: tableau.com ivu (local)

42 abstract 2d/3d data data: GEOMETRIC STRUCTURE MPG Cylinders Horsepower Weight Acceleration Year Origin

43 multivariable visualization how many variables can you visualize? as many as you want! how many variables can you understand? 3? 4?

44 [ Collin Ware, Quantitative Texton Sequences for Legible Bivariate Maps, IEEE, 2009 ] Collin Ware study: encoding 4 variables i.e. encoding 2 variables on map (x,y,u,v) Qton: putative units of pre-attentive human texture perception, analogous to a phoneme in speech recognition (color salad, hold the qtons!)

45

46

47

48

49

50

51 error: integral separable

52

53 reading order/focal flow what do you see first? second? why is it important? what does unknown order create?

54

55

56 science (data analysis, visual analytics) + graphics/hci + graphic design/art (human perception, aesthetics) visualization

57 visualization = science + computer graphics/hci + graphic design/art

58 quick looks at

59

60 ay/bi199: methods of computational science visualization jumpstart+tools+techniques santiago v lombeyda center for advanced computing research caltech