2006 | OriginalPaper | Chapter
Constrained Optimization of the Stress Function for Multidimensional Scaling
Author : Vydunas Saltenis
Published in: Computational Science – ICCS 2006
Publisher: Springer Berlin Heidelberg
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
Multidimensional Scaling (MDS) requires the multimodal Stress function optimization to estimate the model parameters, i.e. the coordinates of points in a lower-dimensional space. Therefore, finding the global optimum of the Stress function is very important for applications of MDS. The main idea of this paper is replacing the difficult multimodal problem by a simpler unimodal constrained optimization problem. A coplanarity measure of points is used as a constraint while the Stress function is minimized in the original high-dimensional space. Two coplanarity measures are proposed. A simple example presented illustrates and visualizes the optimization procedure. Experimental evaluation results with various data point sets demonstrate the potential ability to simplify MDS algorithms avoiding multidimodality.