本体相似度计算和本体映射被广泛应用于查询扩展和图像检索中,已成为信息科学研究的热点内容,其核心为计算本体图中顶点间的相似度.本文从理论的角度分析最优ROC曲线标准下k-部排序本体算法的性质,给出算法模型的单调性、广义误差、可微性等若干统计特征.
Ontology Similarity computing and ontology mapping are widely used in query expansion and image retrieval.The key task ontology applications is calculating the similarity between vertices in ontology graph,and it has become a hot topic in information science research.In this paper,we give some analyses on k-partite ranking ontology algorithm based on optimal ROC curve criterion from a theoretical point of view.Some algorithm statistics characters,such as monotonicity,generalization bound and differentiability,are given.