Model and axioms for similarity metrics

Modern applications usually combine different similarity metrics taking into account the algorithms complexity, the peculiarities of human perception, data resources and samples. The optimization requires a unified formal description of the basic similarity metrics.
The system of the similarity metric axioms is enchanced and its universal model is constructed which generalizes known models of similarity that do not reduce to the Euclidean metric.    The model is based on a weighted partially ordered set. (In Russian).