ann report最近n个点【DOC精选】.docxVIP

? Brief answer The brief answer to your question of that how much is the ratio of computing the nearest neighbors compare to the brutal force computation. From the test results of mine show that this library is highly intelligent, the user can input a ε as the error bound. If the user set the ε value to 0 which means the result of k approximate nearest neighbors search would be exactly identical to the brutal force search for the k nearest neighbors. And due to ANN search build kd-tree or bd-tree as data structure to store points even if we set ε to 0 the search speed will still faster than brutal force search for k nearest neighbors. In nearest neighbor searching, we are given a set of data points S in real d-dimensional space, and are to build a data structure kd-tree or bd-tree. For this application an error bound ε≥0 is also given. The search algorithm returns k distinct points of S, such that for 1≤i ≤k , the ratio between the distances to the ith reported point and the true ith nearest neighbor is at most 1+ ε . Preprocessing time and space are both linear in the number of point n and the dimension d, and are independent of ε. Thus the data structure requires storage that is only moderately larger than the underlying data set. This program can automatically generate datasets; we can set the dimension and how many points we want. Of course we can also input datasets we want to test; each point in d-dimensional space is assumed to be expressed as a d-vector of coordinates p= (p0, p1, …, pd-1) . ANN defines the distance between two points to be their Euclidean distance that is distp,q= (0≤id(pi- qi)2)12 For the purposes of comparing distances, it is not necessary to compute the final square root in the above expression. ANN actually computes squared distances rather than true Euclidean distances. ANN not only saves on the time of computing square roots, but