| Kamalakar Karlapalem
is a professor at IIT, Gandhi Nanagr. Prior to joining IIT,
he was a head of Centre for Data Engineering at International
Institute of Information Technology, Hyderabad. His research
spans the areas of database visualization, data analytics,
workflow management systems, multi-agent systems, and database
systems. He with his students have been awarded in academic
competitions such as, RoboCup, VAST, and TAC. He has graduated
eight PhD and over thirty Masters by Research students.
He is an alumnus of
Indian Statistical Institute (M.Stat), IIT, Kharagpur (M.Tech)
and Georgia Tech (Ph.D) and was a faculty member in the computer
science department at HKUST (1992-2000), before joining IIIT,
Hyderabad.
|
|
Title:
Towards Visualizing Clusters and Classes for Real Valued High
Dimensional Data Sets
Abstract:
High dimensional real valued data sets are the most difficult
to process and mine. Data mining such data sets usually done
by techniques such as, clustering and classification. A major
challenge is to conceptualize the results of the clustering
and classification algorithms. Data visualization helps to
comprehend and get a deeper insight of the data and the data
mining results. The problem is to come up with techniques
and tools to visualize such data and help data scientist to
visualize such data sets and data mining results.
We have built
four tools to visualize and comprehend high dimensional real
valued data. Heidi helps in visualizing the subspace overlap
among the clusters. Beads helps in visualizing spatial spread,
size and shape of the clusters. PEARLS is a visual tool kit
to simultaneous query and explore data sets using concepts
behind Beads. CROVHD helps to visualize spread of data across
dimensions, and to demonstrate closeness and separation of
classes.
In this talk,
I shall present the background for data visualization, technical
insights behind above tools, and list some open problems..
http://www.iiit.ac.in/people/faculty/kamal
|
