Quantization to speedup approximate nearest neighbor search

Abstract The quantization-based approaches not only are the effective methods for solving problems of approximate nearest neighbor search, but also effectively reduce storage space. However, many usually employ fixed nprobes to search process each query. This will lead extra query consumption. Additionally, we observed that as number points in cluster center product quantization increases, cost increases. To address this issue, propose an acceleration strategy based on IVF-HNSW framework further speed up process. involves introducing adaptive termination condition queries and reducing data accessed by building HNSW results. Through extensive experiments, have shown our proposed method significantly accelerates