Thick boundaries in binary space and their influence on nearest-neighbor search

被引:19
作者
Trzcinski, Tomasz [1 ]
Lepetit, Vincent [1 ]
Fua, Pascal [1 ]
机构
[1] Ecole Polytech Fed Lausanne, Comp Vis Lab, CH-1015 Lausanne, Switzerland
关键词
Approximate nearest neighbor search; Binary vectors; Locality sensitive hashing; Hierarchical k-means; ALGORITHM;
D O I
10.1016/j.patrec.2012.08.006
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Binary descriptors allow faster similarity computation than real-valued ones while requiring much less storage. As a result, many algorithms have recently been proposed to binarize floating-point descriptors so that they can be searched for quickly. Unfortunately, even if the similarity between vectors can be computed fast, exhaustive linear search remains impractical for truly large databases and approximate nearest neighbor (ANN) search is still required. It is therefore surprising that relatively little attention has been paid to the efficiency of ANN algorithms on binary vectors and this is the focus of this paper. We first show that binary-space Voronoi diagrams have thick boundaries, meaning that there are many points that lie at the same distance from two random points. This violates the implicit assumption made by most ANN algorithms that points can be neatly assigned to clusters centered around a set of cluster centers. As a result, state-of-the-art algorithms that can operate on binary vectors exhibit much lower performance than those that work with floating point ones. The above analysis is the first contribution of the paper. The second one is two effective ways to overcome this limitation, by appropriately randomizing either a tree-based algorithm or hashing-based one. In both cases, we show that we obtain precision/recall curves that are similar to those than can be obtained using floating point number calculation, but at much reduced computational cost. (c) 2012 Elsevier B.V. All rights reserved.
引用
收藏
页码:2173 / 2180
页数:8
相关论文
共 50 条
[31]   Quantization to speedup approximate nearest neighbor search [J].
Hao Peng .
Neural Computing and Applications, 2024, 36 :2303-2313
[32]   Accelerating Nearest Neighbor Search on Manycore Systems [J].
Cayton, Lawrence .
2012 IEEE 26TH INTERNATIONAL PARALLEL AND DISTRIBUTED PROCESSING SYMPOSIUM (IPDPS), 2012, :402-413
[33]   Highly flexible nearest-neighbor-search associative memory with integrated k nearest neighbor classifier, configurable parallelism and dual-storage space [J].
An, Fengwei ;
Mihara, Keisuke ;
Yamasaki, Shogo ;
Chen, Lei ;
Mattausch, Hans Juergen .
JAPANESE JOURNAL OF APPLIED PHYSICS, 2016, 55 (04)
[34]   Nearest-neighbor and bilinear resampling factor estimation to detect blockiness or blurriness of an image [J].
Suwendi, A ;
Allebach, JP .
DIGITAL PUBLISHING, 2006, 6076
[35]   Accelerating exact nearest neighbor search in high dimensional Euclidean space via block vectors [J].
Zhang, Haowen ;
Dong, Yabo ;
Xu, Duanqing .
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS, 2022, 37 (02) :1697-1722
[36]   A 2D NEAREST-NEIGHBOR QUANTUM ARCHITECTURE FOR FACTORING IN POLYLOGARITHMIC DEPTH [J].
Pham, Paul ;
Svore, Krysta M. .
QUANTUM INFORMATION & COMPUTATION, 2013, 13 (11-12) :937-962
[37]   Angular Distance-Guided Neighbor Selection for Graph-Based Approximate Nearest Neighbor Search [J].
Jung, Sungjun ;
Park, Yongsang ;
Lee, Haeun ;
Oh, Young H. ;
Lee, Jae W. .
PROCEEDINGS OF THE ACM WEB CONFERENCE 2025, WWW 2025, 2025, :4014-4023
[38]   Nearest-neighbor functions for disordered stealthy hyperuniform many-particle systems [J].
Middlemas, Timothy M. ;
Torquato, Salvatore .
JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT, 2020, 2020 (10)
[39]   Revisiting kd-tree for Nearest Neighbor Search [J].
Ram, Parikshit ;
Sinha, Kaushik .
KDD'19: PROCEEDINGS OF THE 25TH ACM SIGKDD INTERNATIONAL CONFERENCCE ON KNOWLEDGE DISCOVERY AND DATA MINING, 2019, :1378-1388
[40]   An Efficient Exact Nearest Neighbor Search by Compounded Embedding [J].
Li, Mingjie ;
Zhang, Ying ;
Sun, Yifang ;
Wang, Wei ;
Tsang, Ivor W. ;
Lin, Xuemin .
DATABASE SYSTEMS FOR ADVANCED APPLICATIONS, DASFAA 2018, PT I, 2018, 10827 :37-54