SymphonyQG: Towards Symphonious Integration of Quantization and Graph for Approximate Nearest Neighbor Search

TL;DR

SymphonyQG introduces a novel integration of quantization and graph-based methods for approximate nearest neighbor search, significantly improving efficiency and accuracy by refining graph structure and eliminating re-ranking steps.

Contribution

It presents SymphonyQG, a new method that better combines quantization and graph techniques, surpassing previous approaches in speed and accuracy for high-dimensional ANN search.

Findings

Achieves state-of-the-art time-accuracy trade-off on real datasets.

Eliminates the need for re-ranking, reducing memory accesses.

Refines graph structure for more efficient search guided by FastScan.

Abstract

Approximate nearest neighbor (ANN) search in high-dimensional Euclidean space has a broad range of applications. Among existing ANN algorithms, graph-based methods have shown superior performance in terms of the time-accuracy trade-off. However, they face performance bottlenecks due to the random memory accesses caused by the searching process on the graph indices and the costs of computing exact distances to guide the searching process. To relieve the bottlenecks, a recent method named NGT-QG makes an attempt by integrating quantization and graph. It (1) replicates and stores the quantization codes of a vertex's neighbors compactly so that they can be accessed sequentially, and (2) uses a SIMD-based implementation named FastScan to efficiently estimate distances based on the quantization codes in batch for guiding the searching process. While NGT-QG achieves promising improvements over the vanilla graph-based methods, it has not fully unleashed the potential of integrating quantization and graph. For instance, it entails a re-ranking step to compute exact distances at the end, which introduces extra random memory accesses; its graph structure is not jointly designed considering the in-batch nature of FastScan, which causes wastes of computation in searching. In this work, following NGT-QG, we present a new method named SymphonyQG, which achieves more symphonious integration of quantization and graph (e.g., it avoids the explicit re-ranking step and refines the graph structure to be more aligned with FastScan). Based on extensive experiments on real-world datasets, SymphonyQG establishes the new state-of-the-art in terms of the time-accuracy trade-off.