An interactive implementation of nonparametric partitioning in ISPAHAN

ISPAHAN, an interactive system for pattern analysis, is described. It features methods for the supervised and unsupervised design of classification strategies, for the design and display of scatter plots and other mappings from high-dimensional space onto a plane, for feature vector transformations, and other interactive techniques designed to facilitate the analysis and interpretation of data in high-dimensional feature spaces. ISPAHAN may also be thought of as a tool for the acquisition of structured knowledge from feature vector data. Since this tool is designed to be versatile and applicable to a variety of feature vector distributions, parametric as well as nonparametric decision functions have been implemented from the beginning. A recent addition to the arsenal of functions is an interactive version of a nonparametric partitioning technique. The implementation and the advantages of interaction are outlined