通过精练查询空间改善高维数据的相似性查询

提出了一种新的高维查询空间过滤策略,通过将高维数据动态投影到低维的数据空间中,然后在投影空间中对查询空间进一步过滤,精练并快速缩小查询空间.同时提出了一种有效的投影策略基于最大间隔投影,这种投影策略能够提高投影空间的精练能力.而且,设计了一种新的高维索引结构MS-tree,并将新的过滤策略运用于MS-tree的范围查询.实验结果表明,这种查询空间精练策略能够有效的提高索引的性能,降低相似性查询的IO代价和CPU代价.

Improving Similarity Search of Multidimensional Data by Reducing Query Space

To perform the query in a high dimensional query space, a novel filtering strategy is proposed. Projecting the high dimensional data into a low dimensional space and filtering the query space in the projected space, the query space is reduced and shrunk quickly. At the same time, an effective projecting strategy is proposed to enhance the reducibility of low dimensional space. Moreover, a new indexing structure or MS-tree is designed with a new filtering strategy applied to the range query of ML-tree. Experimental results show that reducing query space can improve the indexing performance effectively and reduce the cost for IO and CPU.