Parallel Performance of MDS Dimension Reduction Method

Project Details

Project Lead
Seung-Hee Bae 
Project Manager
Seung-Hee Bae 
Institution
Indiana University, Pervasive Technology Institute  
Discipline
Computer Science (401) 

Abstract

Parallel performance analysis of MDS with parallel implementations.

Intellectual Merit

MDS is an useful tool for data visualization, and the parallel implementation of an MDS algorithm is a critical issue to deal with large-scale data visualization.

Broader Impacts

Since MDS algorithms use pairwise dissimilarity information as an input, it can be used for visualization of abstract data sets as well as data represented by feature vectors.

Scale of Use

If possible, I want to use as many nodes as possible for performance study of Parallel MDS Dimension Reduction