Nonlinear Multidimensional Data Projection and Visualisation

Abstract. Multidimensional data projection and visualisation are becoming increasingly important and have found wide applications in many fields such as decision support, bioinformatics and web/document organisation. Various methods and algorithms have been proposed as either nonparametric or semiparametric approaches. This paper provides an overview of the subject and reviews some recent developments. Relationships among various key methods such as Sammon mapping, Neuroscale, principal curve/surface, SOM, GTM and ViSOM are analysed and their advantages and limitations are highlighted in the context of nonlinear principal component analysis and independent component analysis.

• Type: Conference
• Year: 2003
• Where: IDEAL
• Authors: Hujun Yin
• Keywords: Component Analysis, IDEAL 2003, Multidimensional Data Projection, Nonlinear Principal Component