MÍČ, Vladimír, Jan SEDMIDUBSKÝ and Pavel ZEZULA. CRANBERRY: Memory-Effective Search in 100M High-Dimensional CLIP Vectors. Online. In Oscar Pedreira, Vladimir Estivill-Castro. 16th International Conference on Similarity Search and Applications (SISAP). Cham: Springer, 2023, p. 300-308. ISBN 978-3-031-46993-0. Available from: https://dx.doi.org/10.1007/978-3-031-46994-7_26.
Other formats:   BibTeX LaTeX RIS
Basic information
Original name CRANBERRY: Memory-Effective Search in 100M High-Dimensional CLIP Vectors
Authors MÍČ, Vladimír (203 Czech Republic, guarantor), Jan SEDMIDUBSKÝ (203 Czech Republic, belonging to the institution) and Pavel ZEZULA (203 Czech Republic, belonging to the institution).
Edition Cham, 16th International Conference on Similarity Search and Applications (SISAP), p. 300-308, 9 pp. 2023.
Publisher Springer
Other information
Original language English
Type of outcome Proceedings paper
Field of Study 10200 1.2 Computer and information sciences
Country of publisher Czech Republic
Confidentiality degree is not subject to a state or trade secret
Publication form electronic version available online
WWW URL
Impact factor Impact factor: 0.402 in 2005
RIV identification code RIV/00216224:14330/23:00131529
Organization unit Faculty of Informatics
ISBN 978-3-031-46993-0
ISSN 0302-9743
Doi http://dx.doi.org/10.1007/978-3-031-46994-7_26
Keywords in English approximate similarity searching;high-dimensional data;indexing;filtering;LAION dataset
Tags International impact, Reviewed
Changed by Changed by: doc. RNDr. Jan Sedmidubský, Ph.D., učo 60474. Changed: 5/3/2024 11:29.
Abstract
Recent advances in cross-modal multimedia data analysis necessarily require efficient similarity search on the scales of hundreds of millions of high-dimensional vectors. We address this task by proposing the CRANBERRY algorithm that specifically combines and tunes several existing similarity search strategies. In particular, the algorithm: (1) employs the Voronoi partitioning to obtain a query-relevant candidate set in constant time, (2) applies filtering techniques to prune the obtained candidates significantly, and (3) re-rank the retained candidate vectors with respect to the query vector. Applied to the dataset of 100 million 768-dimensional vectors, the algorithm evaluates 10NN queries with 90% recall and query latency of 1.2s on average, all with a throughput of 15 queries per second on a server with 56 core-CPU, and 4.7q/sec. on a PC.
PrintDisplayed: 3/9/2024 06:14