A Graphical User Interface for the Exploratory Analysis of High-Dimensional Data Using ISOMAP

The ISOMAP nonlinear dimensionality reduction method of Tenenbaum, de Silva and Langford, was originally implemented in MATLAB by the developers of the algorithm. One of the issues involved with ISOMAP is the need to determine the number of reduced dimensions that best represents the original data. For this purpose, Tenenbaum, de Silva and Langford provide a plot similar to the scree plot in principal component analysis, where the elbow in the curve represents an estimate of the intrinsic dimensionality. However, for many data sets, the elbow is sometimes difficult to see. Thus, it would be useful to have a Graphical User Interface (GUI) that allows one to explore the results of ISOMAP via scatterplots, parallel coordinates, Andrews curves, and other EDA methods to better understand the effects of dimensionality reduction and to determine the minimum number of dimensions to use. This paper is tutorial in nature and demonstrates a MATLAB GUI for the graphical exploratory analysis of the results of ISOMAP.