Taking all positive eigenvectors is suboptimal in classical multidimensional scaling

DOI10.1137/15M102602XzbMATH Open1351.91020arXiv1402.2703OpenAlexW3105139756MaRDI QIDQ2828334FDOQ2828334

Authors: Jeffrey Tsang, Rajesh Pereira 

Publication date: 25 October 2016

Published in: SIAM Journal on Optimization (Search for Journal in Brave)

Abstract: It is hard to overstate the importance of multidimensional scaling as an analysis technique in the broad sciences. Classical, or Torgerson multidimensional scaling is one of the main variants, with the advantage that it has a closed-form analytic solution. However, this solution is exact if and only if the distances are Euclidean. Conversely, there has been comparatively little discussion on what to do in the presence of negative eigenvalues: the intuitive solution, prima facie justifiable in least-squares terms, is to take every positive eigenvector as a dimension. We show that this, minimizing least-squares to the centred distances instead of the true distances, is suboptimal - throwing away positive eigenvectors can decrease the error even as we project to fewer dimensions. We provide provably better methods for handling this common case.

Full work available at URL: https://arxiv.org/abs/1402.2703

Recommendations

• Better initial configurations for metric multidimensional scaling
• The Solution of the Metric STRESS and SSTRESS Problems in Multidimensional Scaling Using Newtons Method
• Extensions of classical multidimensional scaling via variable reduction
• scientific article; zbMATH DE number 756189
• Multidimensional scaling of noisy high dimensional data

zbMATH Keywords

quadratic programmingconvex optimizationdimensionality reductionmultidimensional scalingclassical multidimensional scaling

Mathematics Subject Classification ID

Factor analysis and principal components; correspondence analysis (62H25) Quadratic programming (90C20) One- and multidimensional scaling in the social and behavioral sciences (91C15)

Cites Work

• Modern multidimensional scaling. Theory and applications.
• Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis
• Remarks to Maurice Frechet's article ``Sur la definition axiomatique d'une classe d'espaces vectoriels distancies applicables vectoriellement sur l'espace de Hilbert
• Some properties of clasical multi-dimesional scaling
• The analytical solution of the additive constant problem
• Interior path following primal-dual algorithms. II: Convex quadratic programming
• Title not available (Why is that?)
• Euclidean distance matrices and applications
• Title not available (Why is that?)

Cited In (7)

• A convex matrix optimization for the additive constant problem in multidimensional scaling with application to locally linear embedding
• A dual basis approach to multidimensional scaling
• Goodness-of-fit filtering in classical metric multidimensional scaling with large datasets
• An analysis of classical multidimensional scaling with applications to clustering
• Horseshoes in multidimensional scaling and local kernel methods
• Central limit theorems for classical multidimensional scaling
• Multidimensional scaling of noisy high dimensional data

Uses Software

• Matlab
• R

This page was built for publication: Taking all positive eigenvectors is suboptimal in classical multidimensional scaling

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2828334)