Research map / Architectures and efficiency / Attention and transformers
Gauge Equivariant Transformer
Lingshen He, Yiming Dong, Yisen Wang, Dacheng Tao, Zhouchen Lin
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 GET instead of HSN | GET · SHREC shape classification | accuracy (%): 96.1 -> 99.2 | Improved |
| use GET instead of MeshCNN | GET · Human Body Segmentation | accuracy (%): 92.3 -> 92.6 | Improved |
| GET has 1/7 the parameters of HSN | GET · SHREC shape classification | parameters: 78K -> 11K | Improved |
| GET has 1/15 the parameters of MeshCNN | GET · Human Body Segmentation | parameters: 2.28M -> 148K | Improved |
| increase group order N from 3 to 9 (ablation) | GET · Human Body Segmentation | accuracy (%): 91.2 -> 92.6 | Improved |
| increase N from 9 to 11 (ablation) | GET · Human Body Segmentation | accuracy (%): 92.6 -> 92.5 | No clear effect |
| replace gauge-equivariant model with non-equivariant GAT (ablation) | GET · Human Body Segmentation | accuracy (%): 92.6 -> 81.1 | Hurt |
| replace attention with convolution (ablation) | GET · Human Body Segmentation | accuracy (%): 92.6 -> 92.3 | Hurt |
| no rotation data augmentation | GET · SHREC shape classification | accuracy (%) | Improved |
Closest papers by meaning
- Gauge Equivariant Convolutional Networks and the Icosahedral CNN (ICML 2019 · similarity 0.89)
- LieTransformer: Equivariant Self-Attention for Lie Groups (ICML 2021 · similarity 0.87)
- Interpreting Equivariant Representations (ICML 2024 · similarity 0.85)
- Neural Isometries: Taming Transformations for Equivariant ML (NeurIPS 2024 · similarity 0.85)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.85)
- Approximately Equivariant Neural Processes (NeurIPS 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.
