Induction of Similarity Measures for Case Based Reasoning Through Separable Data Transformations

Summary

Inducing similarity measures for the case based reasoning scheme through separable data transformations is considered in this chapter. Particular attention is paid to linear transformations of multidimensional data on visualising planes. Separable linear transformations are based both on solutions of eignevalue problems used in the principal componet analysis or in the discriminant analysis as well as on minimization of the convex and piecewise linear (CPL) criterion functions. The perception and the differential criterion functions belong among others to the CPL family. Such functions give possibility for flexible and efficient designing separable transformations of data sets.