Slide 1

Slide 1 text

Building Interactive Data Visualization Systems: A Tool for Light Curve Exploration Zhe Wang PhD Student Department of Computer Science University of Arizona

Slide 2

Slide 2 text

Interactive Data Visualization • Make the visual representation of information respond to human input • Human-Data Interaction

Slide 3

Slide 3 text

Why interactive data visualization? Raw Data Data Product Data Vis Data Analysis Data Vis

Slide 4

Slide 4 text

Data Exploration • First step before a formal data analysis • Help the user to get familiar with the dataset • Ask questions you may not even have raw data collection data cleaning data exploration Design models & algorithms Data product

Slide 5

Slide 5 text

Ok, let’s build a vis app for light curves.

Slide 6

Slide 6 text

Light Curve Time(MJD) Magnitude 52665.418 15.894 52677.430 15.835 52712.315 15.712 … …

Slide 7

Slide 7 text

Traditional Light Curve Visualization

Slide 8

Slide 8 text

Traditional Light Curve Analysis

Slide 9

Slide 9 text

Interactive Light Curve Visualization

Slide 10

Slide 10 text

Technical Detail HTML+Javascript(D3) JSON Web Application • Platform Independent • Easy Access

Slide 11

Slide 11 text

Next Step?

Slide 12

Slide 12 text

No content

Slide 13

Slide 13 text

Real-time Interactive PCA

Slide 14

Slide 14 text

Problem 1: Speed up PCA Calculation • Data Cube • Precompute the intermediate parameters that will be used by PCA • Example (http://vis.stanford.edu/projects/immens/)

Slide 15

Slide 15 text

Speed up PCA calculation • Covariance Matrix a b c … … … extra space for each row in the data cube for d-dimension dataset

Slide 16

Slide 16 text

Problem 2: Speed up PCA Plotting • 1 million points??

Slide 17

Slide 17 text

Speed up PCA Plotting • 1 million points?? Heat map !

Slide 18

Slide 18 text

Speed up PCA Plotting • Specialized Data Structure (For extremely large dataset) 10 20 (5, 20) (5, 15) (15, 20) (5, 10) (10, 15)

Slide 19

Slide 19 text

Demo

Slide 20

Slide 20 text

Technical Detail HTML+Javascript(D3) JSON Client Server Python(Flask) Data Cube

Slide 21

Slide 21 text

Build an interactive visualization system • Overview of the entire dataset + linked charts showing details of subsets. • SVG or Canvas or WebGL? • Special Data Structure

Slide 22

Slide 22 text

Thanks!