Метрические свойства разбиений множеств произвольной природы

Abstract

The interpretation of data content is closely connected with partition analysis. Different applications require different detailings of data partitions. For a system to be successful in a variety of problems, several partitions have to be ensured for cognitive-like techniques. A rational combination of low-level and high-level capabilities seems to be the most promising way to significantly improve the data understanding integrally. To reduce the gap between low-level features and high-level semantics in clustering, we propose, ground, and explore a new metric on partitions of an arbitrary measurable set.