pHNSW: PCA-Based Filtering to Accelerate HNSW Approximate Nearest Neighbor Search

2026-02-22

AI Summary

This work addresses the computational and memory bottlenecks of the Hierarchical Navigable Small World (HNSW) algorithm in high-dimensional approximate nearest neighbor search, which suffers from high computational overhead, irregular memory access patterns, and substantial bandwidth demands. To overcome these challenges, the authors propose a co-optimized algorithm-hardware solution: at the algorithmic level, they introduce PCA-based dimensionality reduction as a filtering step—applied for the first time in this context—to significantly reduce distance computations and graph traversal costs; at the hardware level, they design a dedicated pHNSW processor featuring a custom instruction set and interfaces to both DDR4 and HBM1.0 memory. Implemented in 65nm CMOS technology, the proposed system achieves 14.47–21.37× higher queries per second (QPS) on CPU and 5.37–8.46× on GPU compared to standard HNSW, while reducing energy consumption by up to 57.4%.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Algorithm Configuration
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

Hierarchical Navigable Small World (HNSW) has demonstrated impressive accuracy and low latency for high-dimensional nearest neighbor searches. However, its high computational demands and irregular, large-volume data access patterns present significant challenges to search efficiency. To address these challenges, we introduce pHNSW, an algorithm-hardware co-optimized solution that accelerates HNSW through Principal Component Analysis (PCA) filtering. On the algorithm side, we apply PCA filtering to reduce the dimensionality of the dataset, thereby lowering the volume of neighbor access and decreasing the computational load for distance calculations. On the hardware side, we design the pHNSW processor with custom instructions to optimize search throughput and energy efficiency. In the experiments, we synthesized the pHNSW processor RTL design with a 65nm technology node and evaluated it using DDR4 and HBM1.0 DRAM standards. The results show that pHNSW boosts Queries per Second (QPS) by 14.47x-21.37x on a CPU and 5.37x-8.46x on a GPU, while reducing energy consumption by up to 57.4% compared to standard HNSW implementation.

Problem

• HNSW
• approximate nearest neighbor search
• high-dimensional data
• search efficiency
• computational overhead

Innovation

• PCA filtering
• HNSW acceleration
• algorithm-hardware co-design
• approximate nearest neighbor search
• energy-efficient processor