本体稀疏向量迭代计算算法

Iterative Computation Algorithm for Ontology Sparse Vector

用机器学习方法得到本体函数时,需要将每个本体概念所对应的信息用一个p维向量来表示。在很多应用背景下,由于p值很大而导致计算量庞大。通过稀疏向量的学习来得到本体函数,利用块方法设计一种迭代计算算法进而得到稀疏向量。将该算法应用于生物基因GO本体和物理教育本体,并将实验结果与已有算法的结果作对比,验证了算法在生物基因领域的相似度计算和在物理教育学领域建立本体映射上有较高的效率。

The information for each vertex should be expressed as a p dimension vector when we apply machine learning technology to ontology. In many applications, it leads to large computation cost since p is large. In this paper, the ontology function was obtained in terms of sparse vector learning. The optimal sparse vector was yielded via iterative computation algorithm based on block technology. The algorithm was applied to the GO and physical education ontologies, and the results from our algorithm were compared with results from previous algorithms. It showed that the new algorithm had higher efficiency for calculating the similarity in biology gene field and for establishing the ontology mappings in physical education application.