Patience in Proximity: A Simple Early Termination Strategy for HNSW Graph Traversal in Approximate k-Nearest Neighbor Search

Tommaso Teofili, Jimmy Lin

Abstract: The Hierarchical Navigable Small World (HNSW) graph is widely recognized for its state-of-the-art performance in approximate k-NN search, leveraging a multi-layer proximity graph to efficiently navigate high-dimensional spaces. However HNSW graph traversal can become computationally expensive, especially for large-scale datasets. To address this challenge, we introduce a strategy to early terminate HNSW graph traversal, dubbed “Patience in Proximity”. Inspired by techniques from clustering-informed approximate nearest neighbor algorithms, our approach employs a saturation-based threshold to dynamically halt graph exploration, reducing computational overhead without significantly compromising accuracy.

External IDs:dblp:conf/ecir/TeofiliL25