[论文解读] AQR-HNSW: Accelerating Approximate Nearest Neighbor Search via Density-aware Quantization and Multi-stage Re-ranking

Ganap Ashit Tewary, Nrusinga Charan Gantayat+1 more|arXiv (Cornell University)|Feb 25, 2026

Advanced Image and Video Retrieval Techniques被引用 0

一句话总结

AQR-HNSW 将密度感知自适应量化、多状态重新排序和 SIMD 优化实现结合起来，加速基于 HNSW 的最近邻搜索，在提高吞吐量的同时保持高召回率并降低内存使用。

ABSTRACT

Approximate Nearest Neighbor (ANN) search has become fundamental to modern AI infrastructure, powering recommendation systems, search engines, and large language models across industry leaders from Google to OpenAI. Hierarchical Navigable Small World (HNSW) graphs have emerged as the dominant ANN algorithm, widely adopted in production systems due to their superior recall versus latency balance. However, as vector databases scale to billions of embeddings, HNSW faces critical bottlenecks: memory consumption expands, distance computation overhead dominates query latency, and it suffers suboptimal performance on heterogeneous data distributions. This paper presents Adaptive Quantization and Rerank HNSW (AQR-HNSW), a novel framework that synergistically integrates three strategies to enhance HNSW scalability. AQR-HNSW introduces (1) density-aware adaptive quantization, achieving 4x compression while preserving distance relationships; (2) multi-state re-ranking that reduces unnecessary computations by 35%; and (3) quantization-optimized SIMD implementations delivering 16-64 operations per cycle across architectures. Evaluation on standard benchmarks demonstrates 2.5-3.3x higher queries per second (QPS) than state-of-the-art HNSW implementations while maintaining over 98% recall, with 75% memory reduction for the index graph and 5x faster index construction.

研究动机与目标

• 在生产环境中推动 HNSW 面向十亿级嵌入的可扩展性。
• 开发一种密度感知的自适应量化方案，在不扭曲距离关系的前提下压缩索引。
• 引入多状态重新排序机制以减少不必要的距离计算。
• 设计 SIMD 优化的量化例程，在不同体系结构上最大化硬件吞吐量。

提出的方法

• 实现四倍压缩同时保持距离关系的密度感知自适应量化。
• 通过多状态重新排序将不必要的 ANN 计算减少 35%。
• 量化优化的 SIMD 实现，在不同架构下实现每个周期 16-64 次操作。

实验结果

研究问题

• RQ1 密度感知量化是否能在不损害召回的前提下把 HNSW 索引压缩约 4 倍？
• RQ2 多状态重新排序是否能显著减少 ANN 搜索中的距离计算？
• RQ3 在基于 HNSW 的 ANN 搜索中，使用 SIMD 优化量化能实现多少吞吐量和内存改进？
• RQ4 相对于在标准基准上最先进的 HNSW，在 QPS 和召回方面的相对提升是多少？

主要发现

• 相比最先进的 HNSW 实现，在保持召回率>98%的前提下，查询每秒（QPS）提高 2.5–3.3 倍。
• 索引图的内存减少 75%。
• 索引构建速度提升 5 倍。