Research map / Learning theory and algorithms / Prediction and decision losses
Falconn++: A Locality-sensitive Filtering Approach for Approximate Nearest Neighbor Search
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 Falconn++ instead of Hnsw for indexing | ANNS index · NYTimes | indexing time: 7.8 mins -> 0.6 mins | Helped |
| Use Falconn++ instead of Hnsw for indexing | ANNS index · Glove300 | indexing time: 26.7 mins -> 5.4 mins | Helped |
| Use Falconn++ instead of Hnsw for indexing | ANNS index · Glove200 | indexing time: 13.7 mins -> 1.1 mins | Helped |
| Add locality-sensitive filtering to Falconn | Falconn on Glove200 · Glove200 | query time at the same recall | Helped |
| Reduce the bucket scaling fraction alpha (ablation) | Falconn++ on Glove200 · Glove200 | recall-time tradeoff | Helped |
| Set alpha to 0.1 and add index probing with iProbes=3 (ablation) | Falconn++ on Glove200 · Glove200 | recall-time tradeoff | Helped |
| Increase index probing beyond iProbes=3 (ablation) | Falconn++ on Glove200 · Glove200 | recall-time tradeoff | Hurt |
| Add centering of data points (ablation) | Falconn++ on Glove200 · Glove200 | recall-speed tradeoff | Helped |
| Add limit scaling that keeps at least k points per bucket (ablation) | Falconn++ on Glove200 · Glove200 | recall-speed tradeoff: 0.1n -> 2.42n | Helped |
| Add centering before indexing (ablation) | Falconn on Glove200 · Glove200 | recall-speed tradeoff | Helped |
| Use large iProbes for multi-probe indexing (ablation) | Falconn++ · Glove200 | recall-time tradeoff | Hurt |
| Use the standard limit max(k, alpha B / iProbes) with small k (ablation) | Falconn++ · Glove200, k = 1, 5, 10 | recall-speed tradeoff | Hurt |
Closest papers by meaning
- Practical Near Neighbor Search via Group Testing (NeurIPS 2021 · similarity 0.90)
- Sublinear Time Nearest Neighbor Search over Generalized Weighted Space (ICML 2019 · similarity 0.89)
- Fast k-Nearest Neighbour Search via Prioritized DCI (ICML 2017 · similarity 0.88)
- Graph-Based Algorithms for Diverse Similarity Search (ICML 2025 · similarity 0.86)
- Distance Adaptive Beam Search for Provably Accurate Graph-Based Nearest Neighbor Search (NeurIPS 2025 · similarity 0.86)
- Approximate Nearest Neighbor Search with Window Filters (ICML 2024 · 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.