Research map / Label-efficient and robust learning / Domain adaptation and generalization
Co-Domain Symmetry for Complex-Valued Deep Learning
Utkarsh Singhal, Yifei Xing, Stella X. Yu
Check 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 co-domain symmetric complex layers instead of DCN layers | CIFAR10 and CIFAR100 classifiers · CIFAR10/CIFAR100, RGB, LAB, and Sliding encodings | accuracy: 65.17 -> 69.23; 58.64 -> 67.17; 63.83 -> 69.19; 32.52 -> 41.83; 27.36 -> 39.55; 28.87 -> 42.08 | Improved |
| Use co-domain symmetric complex layers instead of DCN layers | SVHN classifier, Type-I model · SVHN RGB and LAB | accuracy: 85.26 -> 89.39; 84.43 -> 88.86 | Improved |
| Use co-domain symmetric complex layers instead of DCN layers | SVHN classifier, Type-E model · SVHN RGB and LAB | accuracy: 85.26 -> 77.19; 84.43 -> 74.21 | Hurt |
| Use co-domain symmetric complex layers instead of the strongest real-valued or DCN baseline | CIFARnet classifier · SVHN Sliding | accuracy: 87.44 -> 90.25 | Improved |
| Use co-domain symmetric complex layers instead of the strongest real-valued or DCN baseline | CIFARnet classifier · SVHN Sliding | accuracy: 87.44 -> 88.39 | Improved |
| Use co-domain symmetric complex layers instead of SurReal | MSTAR classifier · MSTAR 5% training data | accuracy: 61.1 -> 69.5 | Improved |
| Use co-domain symmetric complex layers instead of SurReal | MSTAR classifier · MSTAR 10% training data | accuracy: 68.0 -> 78.3 | Improved |
| Use co-domain symmetric complex layers instead of SurReal | MSTAR classifier · MSTAR 50% training data | accuracy: 90.3 -> 91.3 | Improved |
| Use co-domain symmetric complex layers instead of SurReal | MSTAR classifier · MSTAR 90% training data | accuracy: 95.6 -> 95.2 | Hurt |
| Use co-domain symmetric complex layers instead of SurReal | MSTAR classifier · MSTAR 100% training data | accuracy: 94.9 -> 96.1 | Improved |
| Use co-domain symmetric complex layers instead of DCN at matched model size | CIFAR10 LAB classifier · CIFAR10 LAB | accuracy: 92.8 -> 93.7 | Improved |
| Use the Type-E composition rather than Type-I (ablation) | Co-domain symmetric CIFAR classifiers · CIFAR100 RGB/LAB/Sliding versus SVHN RGB/LAB/Sliding | accuracy: 36.92 -> 41.83; 37.81 -> 39.55; 38.51 -> 42.08; 89.39 -> 77.19; 88.86 -> 74.21; 90.25 -> 88.39 | Mixed |
Closest papers by meaning
- On Making $SE(2)$-Invariant Networks Optimal (NeurIPS 2026 · similarity 0.83)
- On the hardness of learning under symmetries (ICLR 2024 · similarity 0.83)
- Scale Equivariant Graph Metanetworks (NeurIPS 2024 · similarity 0.82)
- How DNNs break the Curse of Dimensionality: Compositionality and Symmetry Learning (ICLR 2025 · similarity 0.82)
- Equivariant Machine Learning on Graphs with Nonlinear Spectral Filters (NeurIPS 2024 · similarity 0.82)
- Hide & Seek: Transformer Symmetries Obscure Sharpness & Riemannian Geometry Finds It (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.
