Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-Scale Representation Learning for Spatial Feature Distributions using Grid Cells
Gengchen Mai, Krzysztof Janowicz, Bo Yan, Rui Zhu, Ling Cai, Ni Lao
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 multi-scale Space2Vec theory encoding rather than RBF encoding | POI location modeling · Yelp Las Vegas test | MRR: 0.123 (0.001) -> 0.144 (0.001) | Improved |
| Use multi-scale grid-cell representations rather than Adaptive Kernel priors | Fine-grained image classification · BirdSnap† | accuracy: 78.65 -> 79.44 | Improved |
| Use multi-scale theory encoding rather than Adaptive Kernel priors | Fine-grained image classification · NABirds† | accuracy: 81.11 -> 81.59 | Improved |
| Use theory encoding with λmin=50 rather than λmin=1k (ablation) | POI location modeling · Yelp Las Vegas test | MRR: 0.121 (0.001) -> 0.144 (0.001) | Improved |
| Use theory encoding with λmin=50 rather than λmin=500 (ablation) | POI location modeling · Yelp Las Vegas test | MRR: 0.129 (0.001) -> 0.144 (0.001) | Improved |
| Use the full theory encoding rather than block-diagonal constraints (ablation) | POI location modeling · Yelp Las Vegas test | MRR: 0.094 (0.002) -> 0.144 (0.001) | Improved |
| Use theory encoding rather than hexa encoding (ablation) | POI location modeling · Yelp Las Vegas test | MRR: 0.105 (0.001) -> 0.144 (0.001) | Improved |
| Use theory encoding with λmin=10 rather than λmin=200 (ablation) | POI spatial context modeling · Yelp Las Vegas test | MRR: 0.170 (0.001) -> 0.185 (0.002) | Improved |
| Add spatial relationship encoding to context attention (ablation) | Space2Vec context decoder · Yelp Las Vegas test | MRR: 0.167 (0.006) -> 0.185 (0.002) | Improved |
| Use theory rather than grid encoding (ablation) | Fine-grained image classification · BirdSnap† and NABirds† | accuracy: 79.44 vs 79.35; 81.28 vs 81.59 | Mixed |
Closest papers by meaning
- Textual Supervision Enhances Geospatial Representations in Vision-Language Models (ICML 2026 · similarity 0.88)
- Language Models Represent Space and Time (ICLR 2024 · similarity 0.87)
- TorchSpatial: A Location Encoding Framework and Benchmark for Spatial Representation Learning (NeurIPS 2024 · similarity 0.87)
- Poly2Vec: Polymorphic Fourier-Based Encoding of Geospatial Objects for GeoAI Applications (ICML 2025 · similarity 0.87)
- Learning Place Cell Representations and Context-Dependent Remapping (NeurIPS 2024 · similarity 0.86)
- Geolocation Representation from Large Language Models Are Generic Enhancers for Spatio-Temporal Learning (AAAI 2025 · 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.
