Research map / Architectures and efficiency / CNN design
A Single Architecture for Representing Invariance Under Any Space Group
Cindy Zhang, Elif Ertekin, Peter Orbanz, Ryan P Adams
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 symmetry-adapted Fourier positional encoding to the Transformer (ablation) | Transformer · Materials Project total energy | MAE: 0.220 -> 0.197 | Improved |
| add symmetry-adapted Fourier positional encoding to the Transformer (ablation) | Transformer · Materials Project band gap | MAE: 0.440 -> 0.306 | Improved |
| add symmetry-adapted Fourier positional encoding to the Transformer (ablation) | Transformer · Materials Project bulk modulus | MAE: 0.142 -> 0.082 | Improved |
| add symmetry-adapted Fourier positional encoding to the Transformer (ablation) | Transformer · Materials Project shear modulus | MAE: 0.226 -> 0.158 | Improved |
| use CFT instead of ALIGNN | ALIGNN · Materials Project total energy | MAE: 0.203 -> 0.197 | Improved |
| use CFT instead of ALIGNN | ALIGNN · Materials Project band gap | MAE: 0.235 -> 0.306 | Hurt |
| use CFT instead of Matformer | Matformer · Materials Project total energy | MAE: 0.371 -> 0.197 | Improved |
| use CFT instead of Matformer | Matformer · Materials Project bulk modulus | MAE: 0.074 -> 0.082 | No clear effect |
| use CFT instead of CGCNN | CGCNN · Materials Project band gap | MAE: 0.343 -> 0.306 | Improved |
| use CFT instead of ALIGNN for training speed | ALIGNN · Materials Project efficiency | training time per epoch: 592 sec -> 91 sec | Improved |
| use CFT instead of ALIGNN for inference speed | ALIGNN · Materials Project efficiency | inference time: 451 sec -> 60 sec | Improved |
| use CFT instead of Matformer for training speed | Matformer · Materials Project efficiency | training time per epoch: 266 sec -> 91 sec | Improved |
Closest papers by meaning
- Space Group Equivariant Crystal Diffusion (NeurIPS 2025 · similarity 0.89)
- A Space Group Symmetry Informed Network for O(3) Equivariant Crystal Tensor Prediction (ICML 2024 · similarity 0.89)
- Local-Global Associative Frames for Symmetry-Preserving Crystal Structure Modeling (NeurIPS 2025 · similarity 0.87)
- WyckoffDiff -- A Generative Diffusion Model for Crystal Symmetry (ICML 2025 · similarity 0.87)
- Conformal Crystal Graph Transformer with Robust Encoding of Periodic Invariance (AAAI 2024 · similarity 0.87)
- Space Group Conditional Flow Matching (NeurIPS 2026 · similarity 0.87)
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.
