Research map / Learning theory and algorithms / Prediction and decision losses
Learning Space Partitions for 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 Neural LSH instead of k-means partitions | Nearest-neighbor indexing · SIFT | number of candidates at matched 10-NN accuracy: up to 18% more candidates for k-means on average; up to 34% more for the 0.95-quantiles | Helped |
| Use Neural LSH instead of k-means partitions | Nearest-neighbor indexing · GloVe | number of candidates at matched 10-NN accuracy: up to 117% more candidates for k-means on average; up to 130% more for the 0.95-quantiles | Helped |
| Use Neural LSH instead of k-means partitions | Nearest-neighbor indexing · MNIST | number of candidates at matched 10-NN accuracy | Helped |
| Use Regression LSH instead of recursive 2-means | Tree-based nearest-neighbor indexing · GloVe | number of candidates at matched 10-NN accuracy | Helped |
| Use Regression LSH instead of recursive 2-means | Tree-based nearest-neighbor indexing · SIFT | average number of candidates | No clear effect |
| Use Regression LSH instead of recursive 2-means | Tree-based nearest-neighbor indexing · SIFT | 0.95-quantile number of candidates | Helped |
| Use Regression LSH instead of PCA tree | Tree-based nearest-neighbor indexing · GloVe and SIFT | number of candidates at matched 10-NN accuracy | Helped |
| Use Regression LSH instead of random-projection trees | Tree-based nearest-neighbor indexing · GloVe and SIFT | number of candidates at matched 10-NN accuracy | Helped |
| Use Regression LSH instead of recursive 2-means, PCA tree, or random projections | Tree-based nearest-neighbor indexing · MNIST | number of candidates at matched 10-NN accuracy | Helped |
| Add soft labels with S = 15 instead of using S = 1 (ablation) | Neural LSH · GloVe, 256 bins | average number of candidates | Helped |
| Increase soft-label neighborhood size beyond S = 15 (ablation) | Neural LSH · GloVe, 256 bins | overall accuracy | No clear effect |
| Partition the 10-NN graph instead of the 50-NN graph (ablation) | Neural LSH · GloVe, 50-NN search, 16 bins | average number of candidates | No clear effect |
Closest papers by meaning
- A Multilabel Classification Framework for Approximate Nearest Neighbor Search (NeurIPS 2022 · similarity 0.88)
- Discovering Data Structures: Nearest Neighbor Search and Beyond (NeurIPS 2025 · similarity 0.88)
- Probabilistic Routing for Graph-Based Approximate Nearest Neighbor Search (ICML 2024 · similarity 0.85)
- CSPG: Crossing Sparse Proximity Graphs for Approximate Nearest Neighbor Search (NeurIPS 2024 · similarity 0.85)
- Weighted distance nearest neighbor condensing (ICML 2024 · similarity 0.85)
- Sort Before You Prune: Improved Worst-Case Guarantees of the DiskANN Family of Graphs (ICML 2025 · similarity 0.84)
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.