The Theoretical Analysis of Multi-dividing Ontology Learning by Rademacher Vector

Generally, to express the hierarchical and categorical relationship between concepts, the ontology structure is often expressed as a tree diagram. The multi-dividing ontology learning algorithm makes full use of the characteristics of the tree structure ontology graph, and then plays a role in the engineering field. The main contribution of this paper is to use the Rademacher vector method to perform theoretical analysis on the multi-dividing ontology learning algorithm and obtain the error bound estimate for U-statistic criterion. The main proof tricks are to use the skills of statistical learning theory and McDiarmid’s inequality.