Research map / Architectures and efficiency / RNNs and normalization
An Efficient and Robust Framework for Approximate Nearest Neighbor Search with Attribute Constraint
Mengzhao Wang, Lingwei Lv, Xiaoliang Xu, Yuxiang Wang, Qiang Yue, Jiongkang Ni
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use NHQ joint-pruning composite index instead of AF-then-ANNS | DiskANN · BIGANN100M, Recall@10=0.95 | QPS: 315× | Improved |
| use NHQ instead of AF-then-ANNS | DiskANN · BIGANN100M, Recall@10=0.95 | mean disk I/Os: 1,303 -> 30 | Improved |
| use our edge selection instead of HCNNG/NSG (ablation) | NPG edge selection · SIFT1M, Recall@10=0.9 | speedup: 1.1× | Improved |
| use our edge selection instead of HNSW (ablation) | NPG edge selection · SIFT1M, Recall@10=0.9 | speedup: 1.4× | Improved |
| use our edge selection instead of NGT (ablation) | NPG edge selection · SIFT1M, Recall@10=0.9 | speedup: 25.9× | Improved |
| use our routing strategy instead of plain HNSW routing (ablation) | HNSW index · SIFT1M, Recall@10=0.9 | speedup vs HNSW: 1.00 -> 1.21 | Improved |
| use TOGG routing (ablation) | HNSW index · SIFT1M, Recall@10=0.9 | speedup vs HNSW: 1.00 -> 1.20 | Improved |
| use FINGER routing (ablation) | HNSW index · SIFT1M, Recall@10=0.9 | speedup vs HNSW: 1.00 -> 1.30 | Mixed |
| use HCNNG routing (ablation) | HNSW index · SIFT1M, Recall@10=0.9 | speedup vs HNSW: 1.00 -> 1.02 | No clear effect |
| apply NHQ (Strategy C) to HNSW and NSG (ablation) | HNSW, NSG · SIFT1M | QPS vs Recall@10 | Improved |
| use PQ-based IVFPQ with Strategy A instead of PG (ablation) | IVFPQ · SIFT1M | Recall@10 | Hurt |
| use NPGkgraph instead of HNSW or NSG within NHQ (ablation) | NHQ · SIFT1M and GloVe | QPS vs Recall@10 | Improved |
Closest papers by meaning
- Worst-case Performance of Popular Approximate Nearest Neighbor Search Implementations: Guarantees and Limitations (NeurIPS 2023 · similarity 0.89)
- Approximate Nearest Neighbor Search for Modern AI: A Projection-Augmented Graph Approach (ICML 2026 · similarity 0.88)
- LotusFilter: Fast Diverse Nearest Neighbor Search via a Learned Cutoff Table (CVPR 2025 · similarity 0.88)
- SOAR: Improved Indexing for Approximate Nearest Neighbor Search (NeurIPS 2023 · similarity 0.88)
- Graph-Based Algorithms for Diverse Similarity Search (ICML 2025 · similarity 0.87)
- Approximate Nearest Neighbor Search with Window Filters (ICML 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
