11/19/2013
Explainers
Explainers: Expert Explorations with Crafted Projections
An approach to explore high dimensional data: Organize data according to user-defined concepts Explain user-defined concepts according to the data
Michael Gleicher University of Wisconsin – Madison (on sabbatical at INRIA, Rhone-Alpes)
Give the user control over tradeoffs
Warning: this presentation uses the Bariol font family that you have to pay for. I recommend it – it’s cheap and beautiful
Paris
Munich
Sydney
9,2,6,4,… 4,8,1,3,… New York
Atlanta 7,3,2,7,… 5,2,1,7,…
High Dimensional Data
Tokyo
Jakarta 4,8,1,3,… 9,2,6,4,… Beijing 3,2,5,1,… Boston London San Jose 7,3,2,7,… 3,2,5,1,… 5,2,1,7,… 5,2,1,7,…
Objects
have associated Vectors
Paris
Munich
Sydney
9,2,6,4,… 4,8,1,3,… New York
Atlanta 7,3,2,7,… 5,2,1,7,…
Tokyo
Jakarta 4,8,1,3,… 9,2,6,4,… Beijing 3,2,5,1,… Boston London San Jose 7,3,2,7,… 3,2,5,1,… 5,2,1,7,… 5,2,1,7,…
Projections Functions
map Vectors to Numbers
Produce a new axis or dimension (or view)
Paris
Munich
Sydney
9,2,6,4,… 4,8,1,3,… New York
Paris
Atlanta 7,3,2,7,…
5,2,1,7,… Jakarta
Tokyo
4,8,1,3,… 9,2,6,4,… Beijing 3,2,5,1,… Boston London San Jose 7,3,2,7,… 3,2,5,1,… 5,2,1,7,… 5,2,1,7,…
Munich
Sydney
9,2,6,4,… 4,8,1,3,… New York
Atlanta 7,3,2,7,…
5,2,1,7,… Jakarta
Specification
< User-defined concept orange-ness European-ness like-the-marked-things-ness
Tokyo
4,8,1,3,… 9,2,6,4,… Beijing 3,2,5,1,… Boston London San Jose 7,3,2,7,… 3,2,5,1,… 5,2,1,7,… 5,2,1,7,…
Explainer