|
|
KDD-2000 Sixth ACM SIGKDD International Conference on Visualizing
High Dimensional Datasets and Relations
Alfred
Inselberg
Abstract: Intellectual curiosity and the abundance of important multivariate problems, motivate the quest for multidimensional visualization techniques to augment our 3-D perception and experience. Starting from the early successes of data visualization, like Dr. Snow's dot map in 1854 showing the connection of a cholera epidemic to a water pump, a short review of the development is given.� It leads to guidelines for desirable and attainable properties in such methodologies. With the emphasis being on the visualization of high
dimensional data we focus on Parallel Coordinates; a leading
multidimensional/multivariate visualization methodology for this field. The
mathematical foundations for the display and discovery of multidimensional
relations without loss of information are presented interlaced with a variety
of applications. Several multivariate real datasets (i.e. financial,
manufacturing, process control, trading etc) will be displayed and explored
interactively showing how some unsuspected complex relations were discovered
from visual cues suggested by the picture. The derivation of
algorithms� can also be motivated from
this visualization and is illustrated with examples from Computer Vision and
Collision Avoidance for Air Traffic Control. Then geometrical algorithms for Automatic Knowledge
Discovery are derived and applied to real datasets. These algorithms have low
computational complexity, provide explicit and comprehensible rules, do
dimensionality selection by finding ONLY the parameters containing relevant
information, and order these parameters according to some optimality
criteria. Finally, the power to model and display complex nonlinear relations
is illustrated by obtaining, from data, a model of a real country's economy
and interactively discovering plausible economic policies,
interelationships, competition for the same resources, impact of constraints
downstream, sensitivities as well as do trade-off analysis for Decision
Support. PS. Do not be intimidated by this formal
description. The instructor is also well known for his numerological
anecdotes and palindromic digressions! Biography
of Organizer In ancient times AI received a Ph.D. in Applied Math
and Physics from the Univ. of Illinois (Champaign-Urbana) and then held
academic positions in the USA (Univ. of Ill., UCLA, USC) and abroad. In IBM, where
he did research for several years, he became Senior Technical Corporate Staff
Member (a sought after appellation of dubious value). Subsequently, in 1995
he was elected Senior Fellow in Visualization at the San Diego SuperComputing
Center. He has his own company Multidimensional Graphs Ltd and now teaches at
Tel Aviv University. AI invented (1959) and contributed to the development of
Parallel Coordinates, has several patents, over 70 refereed technical papers,
numerous professional and academic awards, and is now writing a book on
Multidimensional Visualization ... and Hi-Tech entertainment. |
|
