Efficient Sketching and Nearest Neighbor Search Algorithms for Sparse Vector Sets

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Bruch, Sebastian, Nardini, Franco Maria, Rulli, Cosimo, Venturini, Rossano
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912615088783360
author Bruch, Sebastian
Nardini, Franco Maria
Rulli, Cosimo
Venturini, Rossano
author_facet Bruch, Sebastian
Nardini, Franco Maria
Rulli, Cosimo
Venturini, Rossano
contents Sparse embeddings of data form an attractive class due to their inherent interpretability: Every dimension is tied to a term in some vocabulary, making it easy to visually decipher the latent space. Sparsity, however, poses unique challenges for Approximate Nearest Neighbor Search (ANNS) which finds, from a collection of vectors, the k vectors closest to a query. To encourage research on this underexplored topic, sparse ANNS featured prominently in a BigANN Challenge at NeurIPS 2023, where approximate algorithms were evaluated on large benchmark datasets by throughput and accuracy. In this work, we introduce a set of novel data structures and algorithmic methods, a combination of which leads to an elegant, effective, and highly efficient solution to sparse ANNS. Our contributions range from a theoretically-grounded sketching algorithm for sparse vectors to reduce their effective dimensionality while preserving inner product-induced ranks; a geometric organization of the inverted index; and the blending of local and global information to improve the efficiency and efficacy of ANNS. Empirically, our final algorithm, dubbed Seismic, reaches sub-millisecond per-query latency with high accuracy on a large-scale benchmark dataset using a single CPU.
format Preprint
id arxiv_https___arxiv_org_abs_2509_24815
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficient Sketching and Nearest Neighbor Search Algorithms for Sparse Vector Sets
Bruch, Sebastian
Nardini, Franco Maria
Rulli, Cosimo
Venturini, Rossano
Data Structures and Algorithms
Information Retrieval
Machine Learning
Sparse embeddings of data form an attractive class due to their inherent interpretability: Every dimension is tied to a term in some vocabulary, making it easy to visually decipher the latent space. Sparsity, however, poses unique challenges for Approximate Nearest Neighbor Search (ANNS) which finds, from a collection of vectors, the k vectors closest to a query. To encourage research on this underexplored topic, sparse ANNS featured prominently in a BigANN Challenge at NeurIPS 2023, where approximate algorithms were evaluated on large benchmark datasets by throughput and accuracy. In this work, we introduce a set of novel data structures and algorithmic methods, a combination of which leads to an elegant, effective, and highly efficient solution to sparse ANNS. Our contributions range from a theoretically-grounded sketching algorithm for sparse vectors to reduce their effective dimensionality while preserving inner product-induced ranks; a geometric organization of the inverted index; and the blending of local and global information to improve the efficiency and efficacy of ANNS. Empirically, our final algorithm, dubbed Seismic, reaches sub-millisecond per-query latency with high accuracy on a large-scale benchmark dataset using a single CPU.
title Efficient Sketching and Nearest Neighbor Search Algorithms for Sparse Vector Sets
topic Data Structures and Algorithms
Information Retrieval
Machine Learning
url https://arxiv.org/abs/2509.24815