Pivot learning for efficient similarity search

Authors:
Manabu Kimura;Kazumi Saito;Naonori Ueda
Affiliations:
Graduate Schools of Information Science, Nara Institute of Science and Technology, Ikoma, Nara, Japan;NTT Communication Science Laboratories, NTT Corporation, Kyoto, Japan;NTT Communication Science Laboratories, NTT Corporation, Kyoto, Japan
Venue:
KES'07/WIRN'07 Proceedings of the 11th international conference, KES 2007 and XVII Italian workshop on neural networks conference on Knowledge-based intelligent information and engineering systems: Part III
Year:
2007

Citing 6
Cited 0

A new version of the nearest-neighbour approximating and eliminating search algorithm (AESA) with linear preprocessing time and memory requirements

Pattern Recognition Letters
Foundations of statistical natural language processing

Foundations of statistical natural language processing
Data structures and algorithms for nearest neighbor search in general metric spaces

SODA '93 Proceedings of the fourth annual ACM-SIAM Symposium on Discrete algorithms
Searching in metric spaces

ACM Computing Surveys (CSUR)
Near Neighbor Search in Large Metric Spaces

VLDB '95 Proceedings of the 21th International Conference on Very Large Data Bases
Pivot selection techniques for proximity searching in metric spaces

Pattern Recognition Letters

Quantified Score

Hi-index	0.00

Visualization

Abstract

Similarity search, finding objects similar to a given query object, is an important operation in multimedia databases, and has many applications in a wider variety of fields. As one approach to efficient similarity search, we focus on utilizing a set of pivots for reducing the number of similarity calculations between a query and each object in a database. In this paper, unlike conventional methods based on combinatorial optimization, we propose a new method for learning a set of pivots from existing data objects, in virtue of iterative numerical nonlinear optimization. In our experiments using one synthetic and two real data sets, we show that the proposed method significantly reduced the average number of similarity calculations, compared with some representative conventional methods.