Research map / Architectures and efficiency / Attention and transformers
A Hierarchical Spatial Transformer for Massive Point Samples in Continuous Space
Wenchong He, Zhe Jiang, Tingsong Xiao, Zelin Xu, Shigang Chen, Ronald Fick, MILES D MEDINA, Christine Angelini
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 |
|---|---|---|---|
| add neural feature learning to GP (Deep GP vs GP) | GP · Red tide | MSE: 7.42 -> 6.23 | Improved |
| use hierarchical quadtree attention (HST) instead of GNN message passing (Spatial GNN) | Spatial GNN · Red tide | MSE: 5.68 -> 5.25 | Improved |
| replace fixed-graph GNN with all-pair attention | Spatial GNN · Red tide | MSE: 5.68 -> 5.30 | Improved |
| multi-level message passing (MGNO) over spatial GNN | Spatial GNN · Red tide | MSE: 5.68 -> 5.41 | Improved |
| HST vs all-pair transformer | All-pair transformer · Red tide | MSE: 5.30 -> 5.25 | Improved |
| HST vs all-pair transformer | All-pair transformer · Turbidity | MSE: 0.31 -> 0.30 | Improved |
| HST vs all-pair transformer | All-pair transformer · Darcy flow | MSE: 0.10 -> 0.11 | Hurt |
| HST vs Galerkin Transformer | GalerkinTransformer · Darcy flow | MAE: 0.25 -> 0.23 | Mixed |
| HST uncertainty branch vs GP | GP · Red tide UQ | AvU: 0.33 -> 0.54 | Improved |
| HST uncertainty branch vs Spatial GNN | Spatial GNN · Red tide UQ | AvU: 0.40 -> 0.54 | Improved |
| HST uncertainty branch vs Galerkin Transformer | GalerkinTransformer · Red tide UQ | AvU: 0.38 -> 0.54 | Improved |
| quadtree leaf size threshold 50 to 200 | HST · computation time simulation | computation time | Hurt |
Closest papers by meaning
- Spatial Transformer for 3D Point Clouds (TPAMI 2021 · similarity 0.86)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.86)
- Locality-Sensitive Hashing-Based Efficient Point Transformer with Applications in High-Energy Physics (ICML 2024 · similarity 0.84)
- Deep Random Features for Scalable Interpolation of Spatiotemporal Data (ICLR 2025 · similarity 0.84)
- PDE-Transformer: Efficient and Versatile Transformers for Physics Simulations (ICML 2025 · similarity 0.84)
- TULIP: Transformer for Upsampling of LiDAR Point Clouds (CVPR 2024 · 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.
