On the representation of training tables in a K-valuedcode and the construction of empirical regularities

摘要

The problem for the transforming of information given in atraining table in an arbitrary code into a k-valued code,without violating the classification, is discussed. The necessary andsufficient condition for the existence of such a transformation in thepattern recognition problems is the nonintersection of classes.Algorithms for effective coding have been suggested. To thek-valued training tables, constructed in this way, it isconvenient to apply combinatorial logic approach for pattern recognitionproblems. This method may be used also in expert systems for theconstruction of empirical regularities in the properties of the objects