On New Clustering Possibilities of Paired Comparisons

In modern data analysis, it is often necessary to solve a clustering problem when only pairwise comparisons of distances or similarities of set elements are presented. Usually, such a situation arises when studying complex objects, structures, or processes. The configuration of objects in the feature space is determined by the data matrix. In the case of pairwise comparisons, the configuration can be changed. Then the clustering results, both by distances and by similarities, may differ. In this case, a new opportunity arises to study and select the best configuration.