Two neural network methods for multidimensional scaling.
Michiel C.V. Wezel, Joost N. Kok, Walter A. Kosters · 1997
Multidimensional scaling (MDS) embeds points in a Euclidean space given only dissimilarity data. Only very recently MDS has gotten some attention from neural network researchers. We propose two neural network methods for MDS and evaluate them using both artificially generated and real data. Training uses two inputs at a time. 1 Introduction Multidimensional scaling (MDS) is a well-known statistical technique that has been applied successfully to a variety of problems in the past. Generally stated, the classical MDS techniques attempt to find a coordinate representation for various objects between which only dissimilarities are given. Dissimilarities between objects are monotonically related to Euclidean distances between the objects. Various types of MDS procedures and objective functions have been presented in the past. Recently MDS also got some attention from neural network researchers [2, 3]. However, in [3] the focus is on MDS as a dimensionality reduction technique, and the prop...