C.H. Ang and T.C. Tan: "New Linear Node Splitting Algorithm for R-trees", Proceedings 5th SSD Conference, pp.339-349, Berlin, Germany, 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....S trees have many similarities with R tree like access methods [29] performance tuning of S trees can be achieved with approaches followed by existing spatial access methods. There is an analogy between weight minimization in Strees and area minimization in R tree nodes. It has been noticed in [28, 30] for R trees (and it also holds for S trees as verified by our experiments) that the insertion of a new entry is biased towards the node with more entries. This takes place during the invocation of the choose leaf procedure, and can be explained by the fact that the node with more entries has ....

Ang, C. H. and Tan, T. C. (1997) New linear node splitting algorithm for R-trees. Proc. 5th SSD Symp., Berlin, Germany, pp. 339--349.

....Figure 1: Architecture of proposed system. bounding box of all the bounding boxes of the entries of the lower level nodes and ptr is the pointer to the lower level node in the R tree. In our implementation, wehave also optimized the R tree using the linear node splitting algorithm proposed in [1]. The data retrieved from the secondary storage are transformed into a ##### ##### [8] using the Transformation Engine) The scene graph is a hierarchical structure that captures the virtual objects and their features suchaslocality, colors, textures and lightings. To better manage the main ....

C. H. Ang and T. C. Tan. New linear node splitting algorithm for r-trees. In Advances in Spatial Databases, SSD'97, pages 339-349, Berlin, Germany, 1997.

.... method use a database of point data containing 120,127 items varied from 2 to 16 dimensions, and compared against several R tree variants; these are R trees [4] S tree [2] Hilbert R tree [14] HG tree [9] with normal point search, and R tree combined with the linear split method by Ang et al. [3]. A detailed evaluation of these results is described in section 7. A.2 A.3 A.1 A.1.1 A.1.2 A.2 A.2.1 A.1 A A.2.2 A.1.1 A.1.2 A.2.1 A.2.2 A.3.1 A.3.2 A.3.3 A.3 A.3 A.3.2 A.3.1 A.3.3 A A.1 A.2 Figure 1: General structure of R tree 3 R tree, Hilbert R tree, and HG tree The ....

....A texture feature extractor [16] is applied to these images and the first 16 features are used for the experiment. The following indexing methods are used for evaluation in this paper: 1. Hilbert R tree [14] 2. R tree [4] 3. S tree [2] 4. R tree using Ang et al. s new linear split [3]. 5. HG tree [9] with PointSearch function The R tree [4] is one of the most successful R tree variants and it has been used by a popular image query application, QBIC [11] The R tree achieves a better storage utilization and retrieval than R tree with re insertion and improved split ....

[Article contains additional citation context not shown here]

C. H. Ang and C. H. Tan. New linear node splitting algorithm for R-trees. In 5th International Symposium on Large Spatial Databases, pages 339 -- 349, Berlin, Germany, July 1997.

....each bounding box o has been associated with two sets) select the partition that ensures the most even distribution of bounding boxes. In case of a tie, choose the partition with the least overlap. In case of another tie, choose the partition with the least coverage. This method takes linear time (Ang Tan 1997). 6. Hilbert Non packed R tree: This dynamic method is based on ordering the objects on the basis of the Peano Hilbert number (e.g. Samet (1990b) corresponding to their centroid (Kamel Faloutsos 1994) The objects are stored in a B tree. 7. MortonNon packed R tree: This dynamic method ....

Ang, C. H. & Tan, T. C. (1997), New linear node splitting algorithm for R-trees, in M. Scholl & A. Voisard, eds, `Advances in Spatial Databases --- Fifth International Symposium, SSD'97', Berlin, Germany, pp. 339--349. (Also Springer-Verlag Lecture Notes in Computer Science 1262).

.... split the overflowing node, while others try to reinsert some of the objects and nodes from the overflowing nodes (Beckmann et al. 1990) thereby striving for better overall behavior (e.g. reduction in coverage and overlap) Methods of intermediate complexity range from taking linear time (e.g. Ang Tan (1997) and (Guttman 1984) to taking quadratic time (Guttman 1984) Other dynamic methods also make use of the ordering applied to the bounding boxes (using their centroids in our examples) These methods are characterized as non packed. The Morton non packed R tree (White 1982) and the Hilbert ....

Ang, C. H. & Tan, T. C. (1997), New linear node splitting algorithm for R-trees, in M. Scholl & A. Voisard, eds, `Advances in Spatial Databases --- Fifth International Symposium, SSD'97', Berlin, Germany, pp. 339--349. (Also Springer-Verlag Lecture Notes in Computer Science 1262).

.... we use merging to implement bulk insertions as done in the cubetree [45] although our merging process is very different) In addition to the numerous bulk loadingand bulk insertion algorithms proposed for the R tree, there have been several different proposals for improving dynamic insertions [4, 9, 10, 43, 31]. Most have been concerned with improving the quality of the resulting partitioning, at the cost of increased construction time, including the well knownR tree method of Beckmann et al. 10] and the polynomial time optimal node splittingmethods of Becker et al. 9] and Garca, Lopez, and ....

....time, including the well knownR tree method of Beckmann et al. 10] and the polynomial time optimal node splittingmethods of Becker et al. 9] and Garca, Lopez, and Leutenegger [43] In addition, 10] and [43] also introduced heuristics for improving storage utilization. Ang and Tan [4] developed a linear time node splittingalgorithm that they claim produces node splits that are better than the original node splitting algorithms [25] and competitive with that of the R tree. The Hilbert R tree of Kamel and Faloutsos [31] employs the same heuristic as the Hilbert packed R tree ....

C. H. Ang and T. C. Tan. New linear node splitting algorithm for R-trees. In Advances in Spatial Databases --- Fifth International Symposium, SSD'97, M. Scholl and A. Voisard, eds., pages 339-- 349, Berlin, Germany, July 1997. (Also Springer-Verlag Lecture Notes in Computer Science 1262).

.... proposed for spatial data, R trees remain a popular index structure employed by numerous commercial DBMS systems [Inf] Map] The R tree structure was initially proposed by Guttman [Gut84] and various variations and improvements over the original structure have been suggested ever since [BKSS90] [AT97] [TS94] Generally, index structures need to be set up by loading data into them before they can be utilized for query processing. Initially, the basic insert operation proposed in [Gut84] was used to load sets of data into an R tree, i.e. each object was inserted one by one into the index ....

C.H. Ang and T.C. Tan. New Linear Node Splitting Algorithm for R-trees. Advances in Spatial Databases, pages 339--349, 1997.

....DBMS systems. R trees, which are an extension of B trees, are designed to store multi dimensional data [A.G84] The R tree structure was initially proposed by Guttman [A. G84] and various variations and improvements over the original structure have been suggested and implemented ever since [NHRB90] [CT97] [YT94] Note that like with most other index structures, the data needs to be first loaded into the R tree structure in order to set up the index before the index structure can be exploited for query processing. Initially, the basic insert operation proposed in [A.G84] was used to load sets of ....

....refer to this traditional technique of data loading as the OBO (one by one) technique. The original structure provided by Guttman [A. G84] did not use the best strategies for insertion and more optimal insertion and node splitting techniques were provided subsequently by Theodoridis [YT94] and Ang [CT97]. These new insertion techniques somewhat improved the R tree loading times. The insertion of data sequentially into the R tree may be suitable for situations when we are entering a few data items into the R tree. It has however been found to be extremely inefficient in the case when the entire ....

C.H.Ang and T.C.Tan. New linear node splitting algorithm for r-trees. Advances in Spatial Databases, pages 339--349, 1997.

....DBMS systems. R trees, which are an extension of B trees, are designed to store multi dimensional data [A.G84] The R tree structure was initially proposed by Guttman [A. G84] and various variations and improvements over the original structure have been suggested and implemented ever since [NHRB90] [CT97] [YT94] Note that like with most other index structures, the data needs to be first loaded into the R tree structure in order to set up the index before the index structure can be exploited for query processing. Initially, the basic insert operation proposed in [A.G84] was used to load sets of ....

....R tree, i.e. each object was inserted one by one into the index. In this paper, we will refer to this traditional technique of data loading as the OBO (one by one) technique. Over the last few years, more optimal insertion and node splitting techniques were provided by Theodoridis [YT94] and Ang [CT97]. These new insertion techniques somewhat improved the R tree loading times. The insertion of data sequentially into the R tree may be suitable for situations when we are entering a few data items into the R tree. It has however been found to be inefficient in the case when the entire tree needs ....

C.H.Ang and T.C.Tan. New linear node splitting algorithm for R-trees. Advances in Spatial Databases, pages 339--349, 1997.

....DBMS systems. R trees, which are an extension of B trees, are designed to store multi dimensional data [Gut84] The R tree structure was initially proposed by Guttman [Gut84] and various variations and improvements over the original structure have been suggested and implemented ever since [NHRB90] [CT97] [YT94] Note that like with most other index structures, the data needs to be first loaded into the R tree structure in order to set up the index before the index structure can be exploited for query processing. Initially, the basic insert operation proposed in [Gut84] was used to load sets of ....

....each object was inserted one by one into the index structure. In this paper, we will refer to this traditional technique of data loading as the OBO (one by one) technique. Over the last few years, more optimal insertion and node splitting techniques were provided by Theodoridis [YT94] and Ang [CT97]. These new insertion techniques somewhat improved the R tree loading times. The insertion of data sequentially into the R tree may be suitable for situations when we are entering a few data items into the R tree. It has however been found to be inefficient in the case when the entire tree needs ....

C.H.Ang and T.C.Tan. New linear node splitting algorithm for R-trees. Advances in Spatial Databases, pages 339--349, 1997.

....Figure 1: Architecture of proposed system. bounding box of all the bounding boxes of the entries of the lower level nodes and ptr is the pointer to the lower level node in the R tree. In our implementation, we have also optimized the R tree using the linear node splitting algorithm proposed in [1]. The data retrieved from the secondary storage are transformed into a scene graph [8] using the Transformation Engine) The scene graph is a hierarchical structure that captures the virtual objects and their features such as locality, colors, textures and lightings. To better manage the main ....

C. H. Ang and T. C. Tan. New linear node splitting algorithm for r-trees. In Advances in Spatial Databases, SSD'97, pages 339-349, Berlin, Germany, 1997.

No context found.

C.H. Ang and T.C. Tan: "New Linear Node Splitting Algorithm for R-trees", Proceedings 5th SSD Conference, pp.339-349, Berlin, Germany, 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC