Data Structures and Algorithms for Nearest Neighbor Search in General Metric Spaces：（数据结构和算法的最近邻搜索一般度量空间）.pdfVIP

Data Structures and Algorithms for Nearest Neighb or Search in General Metric Spaces Peter N Yianilos  Abstract nal knowledge Ie it do esnt matter what the database We consider the computational problem of nding nearest ob jects are neighb ors in general metric spaces Of particular interest are So while the notions of ordering and metricdistance spaces that may not b e conveniently emb edded or approxi mated in Euclidian space or where the dimensionality of a are only lo osely analogous we are nevertheless motivated Euclidian representation is very high to lo ok for improved search metho ds that dep end only Also relevant are highdimensional Euclidian settings in on the information received through metric evaluation which the distribution of data is in some sense of lower di mension and emb edded in the space Now binary search presumes that the database has The vptree vantage p oint tree is intro duced in several b een sorted an n logn pro cess We then seek to or forms together with asso ciated algorithms as an improved ganize our database in as much time so that under some metho d for these dicult search problems Tree construc tion executes in Onlog n time and search is under certain circumstances logarithmic time search is p ossible circumstances and in the limit Olog n exp ected time We intro duce the vptree vantage p oint tree in several The theoretical basis for this approach is develop ed and forms as one solution This data structure and the al the results of several exp eriments are rep orted In Euclidian gorithms to build and search it were rst discovered and cases kdtre