Pattern recognition with mixed and incomplete data
José Ruíz-Shulcloper · Pattern Recognition and Image Analysis · 2008
In this paper, an increasingly active line of study in pattern recognition (PR) called “logical combinatorial pattern recognition” (LCPR) is reviewed. Briefly, this refers to pattern recognition problems with mixed and incomplete object descriptions using similarity functions less restricted than a distance, i.e., objects described simultaneously in terms of numerical and non-numerical features with missing values. The similarity function is not necessarily the opposite or the inverse of a certain distance function; it could be asymmetric. The necessity of this branch of PR and the ways in which these problems have been faced, as well as its basic concepts, tools, and the principal theoretical and practical results over the last 30 years and the most important future works, are concisely exposed.