Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs

By Yu. A. Malkov et al

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Works
2.1 Proximity graph techniques
2.2 Navigable small world models
3. Motivation
4. Hierarchical NSW approach
5. Heuristic for selecting graph neighbors

Summary

The document discusses a new approach for approximate nearest neighbor search based on Hierarchical Navigable Small World (HNSW) graphs. The proposed solution is fully graph-based, offering better performance compared to existing approaches. It incrementally builds a multi-layer structure of proximity graphs for nested subsets of stored elements. The algorithm utilizes scale separation and a heuristic for selecting proximity graph neighbors. Performance evaluation shows significant improvement over state-of-the-art vector-only approaches. The Hierarchical NSW algorithm provides a logarithmic complexity scaling and outperforms previous approaches on various datasets.