Presentation 2009-07-28
Maximal Metric Margin Partitioning for Similarity Search Index
Hisashi KURASAWA, Daiji FUKAGAWA, Atsuhiro TAKASU, Jun ADACHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A fundamental issue that confronts the development of an index for similarity searches in metric spaces is how to divide the data effectively for search pruning. We propose Maximal Metric Margin Partitioning (MMMP), a partitioning scheme for similarity search indexes. MMMP divides the data space based on its distribution patterns, especially the boundaries of clusters. A partitioning surface created by MMMP is at maximum distances from the two cluster boundaries. MMMP is the first similarity search index approach to focus on the partitioning surfaces and data distribution patterns. We also present an indexing scheme, named the MMMP-Index, that uses MMMP and pivot filtering. The MMMP-Index discards many objects that are not relevant to a query by MMMP, and it reduces the query execution cost. Experimental results show that MMMP effectively indexes clustered data and reduces the search cost. For clustered vector data, the MMMP-Index reduces the computational cost to less than two thirds that of the compared schemes.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Similarity Search / Index / Metric Space / Margin
Paper # DE2009-3
Date of Issue

Conference Information
Committee DE
Conference Date 2009/7/21(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Data Engineering (DE)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Maximal Metric Margin Partitioning for Similarity Search Index
Sub Title (in English)
Keyword(1) Similarity Search
Keyword(2) Index
Keyword(3) Metric Space
Keyword(4) Margin
1st Author's Name Hisashi KURASAWA
1st Author's Affiliation The University of Tokyo()
2nd Author's Name Daiji FUKAGAWA
2nd Author's Affiliation National Institute of Informatics
3rd Author's Name Atsuhiro TAKASU
3rd Author's Affiliation National Institute of Informatics
4th Author's Name Jun ADACHI
4th Author's Affiliation National Institute of Informatics
Date 2009-07-28
Paper # DE2009-3
Volume (vol) vol.109
Number (no) 153
Page pp.pp.-
#Pages 6
Date of Issue