A study of MBR-based spatial access methods: how well they perform in high-dimensional spaces

Orlandić, Ratko; Yu, Byunggu

doi:10.1109/ideas.2000.880594

Cited by 7 publications

(13 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another problem of the transformation schemes is that they do not scale well to the increasing dimensionality of the data space [23]. One reason for this is that the transformation doubles the size of index entries in the interior nodes of the underlying PAM.…”

Section: Figure 1 Endpoint Transformation Of One-dimensional Segmentsmentioning

confidence: 99%

“…Since the interior nodes of a typical PAM maintain rectangular regions [27,30], twice as many numbers are required to represent a 2d-dimensional than a d-dimensional rectangle. 1 Due to the larger entries, the transformation methods place a tighter limit on the dimensionality of the data space than most other schemes [23].…”

Section: Figure 1 Endpoint Transformation Of One-dimensional Segmentsmentioning

confidence: 99%

“…The magnitude of this problem increases rapidly with a growing number of dimensions d [23]. For example, assume an intersection query with a d-dimensional square query window whose extension along every axis is s > 0.…”

Section: Figure 1 Endpoint Transformation Of One-dimensional Segmentsmentioning

confidence: 99%

“…The main reason for the high improvements over the R * -tree variants is the region overlap in R * -trees, which increases the probability that an index node is accessed. In contrast, since the QSF-tree performs a kdB-tree-like partition of space that incurs no region overlap, it avoids the penalties resulting from the region overlap [23].…”

Section: Object Transformation Vs Other Samsmentioning

confidence: 99%

“…Since this query transformation refines the basic idea of the second-generation object transformation [14,29], we call it the refined second-generation (r2G) transformation scheme. But, it can also be viewed as an extension of the transformation technique applied by our recent design, called QSF-trees [31,23]. In the paper, we choose to emphasize the later point of view as it enables derivation of more compact search regions as well as a formal proof of correctness of the query transformation for arbitrary d-dimensional spaces.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Object and query transformation

Orlandić

2000

Proceedings of the Ninth International Conference on Information and Knowledge Management

View full text Add to dashboard Cite

The complexity of deploying high-performance spatial structures in transactional DBMS environments has motivated researchers to experiment with the idea of reusing the effort invested in developing a traditional indexing scheme. To enable reuse, the design of a spatial access method must employ a form of object transformation that maps extended objects in the original space into points of a dual space. If these points are further transformed into one-dimensional keys, they can be indexed by any traditional scheme for exact-match retrieval, e.g. B + -trees. This paper investigates the performance of two generations of object transformation schemes in both low-and high-dimensional spaces. The first-generation object transformation provides relatively good performance in low-dimensional spaces. But, as the number of dimensions grows, its performance deteriorates quickly. The second generation scheme applies a query transformation that refines an idea, introduced in connection with the design of LSD-trees, of limiting the search space by taking into account properties of the actual data set. The refined secondgeneration scheme enables a rapid deployment of a spatial access method while achieving performance comparable to that of advanced monolithic schemes.

show abstract

Section: Figure 1 Endpoint Transformation Of One-dimensional Segmentsmentioning

confidence: 99%

Section: Figure 1 Endpoint Transformation Of One-dimensional Segmentsmentioning

confidence: 99%

Section: Figure 1 Endpoint Transformation Of One-dimensional Segmentsmentioning

confidence: 99%

Section: Object Transformation Vs Other Samsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations