Research map / Architectures and efficiency / RNNs and normalization
A Simple Hierarchical Pooling Data Structure for Loop Closure
Xiaohan Fei, Konstantine Tsotsos, Stefano Soatto
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 hierarchical mean-pooling instead of flat inverted index | inverted index search · KITTI overall (10K vocab) | speedup: 1.00 -> 26.00 | Improved |
| use hierarchical sum-pooling | inverted index search · KITTI overall (10K vocab) | speedup: 1.00 -> 1.25 | Improved |
| use hierarchical max-pooling | inverted index search · KITTI overall (10K vocab) | speedup: 1.00 -> 1.63 | Improved |
| add max-pooling hierarchy to ORB-SLAM loop closure | ORB-SLAM · KITTI | mean query time (ms): 5.80 -> 2.04 | Improved |
| switch to 1M vocabulary: sum/max-pooling still help | inverted index search · KITTI overall (1M vocab) | speedup: 1.00 -> 2.13 | Improved |
| use mean-pooling with larger vocabulary (ablation) | hierarchical mean-pooling · KITTI (1M vocab) | precision-recall | Hurt |
| search among keyframes only (no hierarchy) (ablation) | inverted index search · KITTI | query time | Hurt |
| replace regular sampling by adaptive spectral clustering in SE(3) (ablation) | hierarchical structure · KITTI 00 and 02 | query time | No clear effect |
| use parallax-based sampling for grouping (ablation) | hierarchical structure · KITTI | query time | No clear effect |
| hierarchical mean-pooling on keyframes (Oxford) | inverted index search · Oxford (10K vocab, keyframes) | speedup: 1.00 -> 4.34 | Improved |
| use hierarchy on ukbench retrieval (random grouping, sum) | inverted index search · ukbench | top-4 score: 2.72 -> 2.83 | Improved |
| use hierarchy on INRIA Holidays retrieval (random grouping, sum) | inverted index search · INRIA Holidays | mAP: 0.56 -> 0.63 | Improved |
Closest papers by meaning
- Pool Me Wisely: On the Effect of Pooling in Transformer-Based Models (NeurIPS 2025 · similarity 0.83)
- Jet-Long: Efficient Long-Context Extension with Dynamic Bifocal RoPE (NeurIPS 2026 · similarity 0.82)
- Hierachical Balance Packing: Towards Efficient Supervised Fine-tuning for Long-Context LLM (NeurIPS 2025 · similarity 0.81)
- Sparse Layers are Critical to Scaling Looped Language Models (NeurIPS 2026 · similarity 0.81)
- Stability and Generalization in Looped Transformers (NeurIPS 2026 · similarity 0.80)
- Slower Generalization, Faster Memorization: A Sweet Spot in Algorithmic Learning (NeurIPS 2026 · similarity 0.80)
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.
