Searching in one billion vectors: re-rank with source coding

TL;DR

This paper introduces a re-ranking method for high-dimensional vector search that refines neighbor hypotheses using source coding techniques, reducing memory usage and improving accuracy in billion-scale datasets.

Contribution

It proposes a novel re-ranking approach based on source coding that enhances existing indexing methods for billion-scale high-dimensional vector search.

Findings

Accurately re-ranks neighbors with minimal memory

Efficiently refines distances using short quantization codes

Demonstrates effectiveness on a new billion-vector dataset

Abstract

Recent indexing techniques inspired by source coding have been shown successful to index billions of high-dimensional vectors in memory. In this paper, we propose an approach that re-ranks the neighbor hypotheses obtained by these compressed-domain indexing methods. In contrast to the usual post-verification scheme, which performs exact distance calculation on the short-list of hypotheses, the estimated distances are refined based on short quantization codes, to avoid reading the full vectors from disk. We have released a new public dataset of one billion 128-dimensional vectors and proposed an experimental setup to evaluate high dimensional indexing algorithms on a realistic scale. Experiments show that our method accurately and efficiently re-ranks the neighbor hypotheses using little memory compared to the full vectors representation.