Exploring the linearity of models on the basis of ranked data
Leon Bobrowski and Ralph Huntsinger
Summer Computer Simulation Conference 2007 (SCSC 2007)
San Diego, California (USA), July 15-18, 2007
Abstract
Abstract: The ranked data are based on two sources of information: on results of measurements that are represented as a set of n-dimensional feature vectors, and on ranking relations between some of these vectors. The ranked data can be modelled in the form of a linear transformation from the feature space on a line. The ranked transformation preserves as much as possible the ranking relations between feature vectors. The linear ranked transformations could be explored on the basis of the linear separability of the differential vectors. A linear transformation can fully reflect the ranking relations if the feature vectors are linearly independent. Key words: linear transformations, ranked relations, convex and piecewise linear (CPL) criterion functions, linear separability, multiparameter optimization