ANALYSIS AND DESCRIPTION OF NON-DIGITAL DATA BY SUBSETS

The methods suggested make it possible to find clasters with the given properties and to optimize sets of solution rules for voting algorithms of recognition. Algorithms, realizing the methods suggested, have been constructed. The notion of the t-stable set has been introduced. Some methods and practical problems for the analysis of data have been represented in the form of the search of t-stable sets, in particular, maximum of a submodular function. The construction problem of a set of partial solutions for the voting has been conversed to maximization of a submodular function. The way of maximization - the method of the distributed teaching has been suggested. The generalization of the hierarchical claster analysis, generalization permitting clasters, crossing on a one level, and incomplete classifications, has been constructed. The optimum algorithm for the search of maximum of a submodular function has been constructed. The paper results may find their field of application in medicine, geology and other fieldsAvailable from VNTIC / VNTIC - Scientific & Technical Information Centre of RussiaSIGLERURussian Federatio