公开(公告)号：KR100419575B1

公开(公告)日：2004-02-19

申请号：KR1020000073539

申请日：2000-12-05

Applicant: 한국전자통신연구원

Inventor： 이훈순 ,  김준 ,  유재수 ,  송석일 ,  피준일

IPC: G06F17/30

CPC classification number: G06F17/30333 ,  Y10S707/99934 ,  Y10S707/99935

Abstract: A bulk loading method, for use in a high-dimensional index structure using some parts of dimensions based on an unbalanced binarization scheme, accelerates an index construction and improves a search performance. For the purpose, the bulk loading method calculates a topology of the index by recognizing information for the index to be constructed using a given data set, splits the given data set into sub-sets of data by repeatedly performing an establishment of a split strategy and a binarization based on the calculated topology of the index, if a leaf node is derived from the sub-sets of data divided through a top-down recursive split process, reflects a minimum bounding region of the leaf node on a higher node, and, if a non-leaf node is generated, repeatedly performing the above processes for another sub-set of data to thereby produce a final root node.

Abstract translation: 批量加载方法用于使用基于不平衡二值化方案的维度的某些部分的高维索引结构，加快了索引构建并提高了搜索性能。 为此目的，批量加载方法通过使用给定数据集识别要构建的索引的信息来计算索​​引的拓扑，通过重复执行分离策略的建立来将给定数据集拆分为数据的子集，并且 如果从通过自顶向下递归分割处理划分的数据子集导出叶节点，则基于所计算的索引拓扑的二值化反映了较高节点上的叶节点的最小边界区域， 如果生成了非叶节点，则对另一个数据子集重复执行上述过程，从而产生最终的根节点。

公开(公告)号：KR1020020044029A

公开(公告)日：2002-06-14

申请号：KR1020000073539

申请日：2000-12-05

Applicant: 한국전자통신연구원

Inventor： 이훈순 ,  김준 ,  유재수 ,  송석일 ,  피준일

IPC: G06F17/30

CPC classification number: G06F17/30333 ,  Y10S707/99934 ,  Y10S707/99935

Abstract: PURPOSE: A bulk loading method for a high dimensional index structure is provided to be suitable to the high dimensional index structure using a partial dimension based on an unbalanced bisection method of the UBBT(Unbalance Bisectional Bulk-loading) and to improve the index configuration time and a search function. CONSTITUTION: The bulk loading method for high dimensional index structure comprises the steps of calculating a format of an index structure by understanding the information for the index to be formed as a given data set, dividing the given data set into the sub data set while repeatedly executing the establishment of the division strategy and the bisection method on a basis of the calculated index structure, reflecting a minimum border area of an end node to an upper level if only one end node is generated from the sub data set by executing the repeated dividing process, and generating a final root node by repeated executing the previous steps for the other sub data set if one non-end node is generated.

Abstract translation: 目的：提供一种适用于高尺寸折射率结构的体积加载方法，适用于使用基于UBBT（不平衡二分批装载）的不平衡二分法的部分维度的高维度索引结构，并提高索引配置时间 和搜索功能。 构成：用于高维度索引结构的批量加载方法包括以下步骤：通过理解要形成为给定数据集的索引的信息来计算索​​引结构的格式，同时重复地将给定的数据集划分为子数据集 在计算出的索引结构的基础上执行划分策略和二分法的建立，如果仅通过执行重复划分从子数据集生成一个端节点，则将端节点的最小边界区域反映到较高级别 过程，并且如果生成了一个非端节点，则通过重复执行其他子数据集的前述步骤来生成最终根节点。