Research map / Architectures and efficiency / RNNs and normalization
Intrinsic Lorentz Neural Network
Xianglong Shi, Ziheng Chen, Yunhan Jiang, Nicu Sebe
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 fully intrinsic ILNN instead of Euclidean ResNet-18 | ResNet-18 · CIFAR-10 / CIFAR-100 | accuracy: 95.14 -> 95.36; 77.72 -> 78.41 | Improved |
| ILNN vs HCNN Lorentz | ILNN · CIFAR-10 / CIFAR-100 | accuracy: 95.14 -> 95.36; 78.07 -> 78.41 | Improved |
| ILNN vs Euclidean CNN and HCNN-S | ILNN · TEB and GUE genomics | MCC: Covid 36.71 (HCNN-S) -> 64.76; tata core 79.54 -> 83.90 | Improved |
| replace Linear layer in Hypformer with PLFC | Hypformer · Airport, Cora, PubMed | accuracy: 95.0 -> 96.0; 85.0 -> 85.7; 81.3 -> 82.5 | Improved |
| add point-to-hyperplane FC (PLFC) instead of LFC (ablation) | ILNN with GyroLBN · CIFAR-10 | accuracy: 95.19 -> 95.36 | Improved |
| add point-to-hyperplane FC (PLFC) instead of LFC (ablation) | ILNN with GyroLBN · GUE core promoter tata | MCC: 81.33 -> 83.90 | Improved |
| use GyroLBN instead of LBN (ablation) | LFC head · GUE core promoter notata | MCC: 66.60 -> 71.92 | Improved |
| use GyroLBN instead of LBN (ablation) | LFC head · CIFAR-10 | accuracy: 95.14 -> 95.19 | No clear effect |
| use GyroLBN instead of GyroBN (ablation) | PLFC head · CIFAR-10 | accuracy: 95.28 -> 95.36 | No clear effect |
| use GyroLBN instead of GyroBN (ablation) | PLFC head · CIFAR-10 fit time per epoch | training time: 314 -> 169 | Improved |
| use GyroLBN instead of LBN (ablation) | LFC head · CIFAR-10 fit time per epoch | training time: 142 -> 125 | Improved |
| use mask-and-project Lorentz dropout instead of log-exp dropout (ablation) | ILNN · unspecified | performance | Improved |
Closest papers by meaning
- Lorentz Local Canonicalization: How to make any Network Lorentz-Equivariant (NeurIPS 2025 · similarity 0.85)
- Fully Hyperbolic Convolutional Neural Networks for Computer Vision (ICLR 2024 · similarity 0.85)
- Robust Hyperbolic Learning with Curvature-Aware Optimization (NeurIPS 2025 · similarity 0.83)
- Clifford-Steerable Convolutional Neural Networks (ICML 2024 · similarity 0.83)
- Lorentz-Equivariant Geometric Algebra Transformers for High-Energy Physics (NeurIPS 2024 · similarity 0.82)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.82)
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.
