Research map / Learning theory and algorithms / Prediction and decision losses
Connecting Compression Spaces with Transformer for Approximate Nearest Neighbor Search
Haokui Zhang, Buzhou Tang, Wenze Hu, Xiaoyu Wang
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 |
|---|---|---|---|
| compress features x4 for HNSW index building | HNSW · GIST1M | indexing time (s): 733 -> 245 | Improved |
| compress features x4 for HNSW index building | HNSW · Deep1M | indexing time (s): 300 -> 100 | Improved |
| compress features x4 for NSG index building | NSG · GIST1M | indexing time (s): 1118 -> 645 | Improved |
| compress features x4 for HNSW index building | HNSW · GIST1M | recall 100@100: 94.39 -> 95.27 | Improved |
| compress features x2 for Bigann-1B index | HNSW · Bigann-1B | indexing time (h): 106 -> 51 | Improved |
| add CCST compression before PQ (60 bytes) | PQ · GIST1M | recall 1@1: 23.0 -> 39.5 | Improved |
| add CCST compression before PQ (32 bytes) | PQ · Deep1M | recall 1@1: 32.3 -> 53.3 | Improved |
| add CCST compression before L&C (60 bytes) | L&C · GIST1M | recall 1@1: 31.4 -> 30.7 | Mixed |
| use CCST instead of Catalyst for compression before PQ (16 bytes) | PQ · Deep1M | recall 1@1: 19.8 -> 32.5 | Improved |
| use CCST instead of VAE for compression x4 | HNSW · GIST1M | recall 1@1: 49.2 -> 80.1 | Improved |
| add scalar quantization on top of CCST x4 (ablation) | HNSW + CCST · GIST1M | recall 100@100: 95.27 -> 91.42 | Hurt |
| add scalar quantization on top of CCST x4 (ablation) | HNSW + CCST · GIST1M | indexing time (s): 245 -> 121 | Improved |
Closest papers by meaning
- Product Quantized Translation for Fast Nearest Neighbor Search (AAAI 2018 · similarity 0.89)
- Distance Encoded Product Quantization for Approximate K-Nearest Neighbor Search in High-Dimensional Space (TPAMI 2018 · similarity 0.88)
- SOAR: Improved Indexing for Approximate Nearest Neighbor Search (NeurIPS 2023 · similarity 0.87)
- LoRANN: Low-Rank Matrix Factorization for Approximate Nearest Neighbor Search (NeurIPS 2024 · similarity 0.86)
- Approximate Nearest Neighbor Search with Window Filters (ICML 2024 · similarity 0.86)
- LotusFilter: Fast Diverse Nearest Neighbor Search via a Learned Cutoff Table (CVPR 2025 · similarity 0.85)
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.
