Research map / Architectures and efficiency / CNN design
Towards an Effective Orthogonal Dictionary Convolution Strategy
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Apply ODCS: remove the intermediate ReLU and regularize the dictionary convolution | ResNet basic-block variants · CIFAR-10 and CIFAR-100 | accuracy: 92.65 -> 93.13; 93.89 -> 94.29; 93.96 -> 94.55; 79.20 -> 79.50; 82.33 -> 82.65; 82.84 -> 83.08 | Helped |
| Apply ODCS to the bottleneck block | ResNet-50 and ResNet-101 · CIFAR-10 and CIFAR-100 | accuracy: 92.98 -> 93.57; 92.76 -> 93.60; 82.90 -> 84.62; 83.51 -> 84.55 | Helped |
| Apply ODCS | ResNet-34, ResNet-50, and ResNet-101 · CUB-200-2011 | accuracy: 81.45 -> 81.93; 81.17 -> 81.55; 82.82 -> 82.96 | Helped |
| Apply ODCS | ResNet-34, ResNet-50, and ResNet-101 · ImageNet-2012 | accuracy: 74.37 -> 76.30; 76.73 -> 77.00; 76.60 -> 76.87 | Helped |
| Remove ReLU between convolutions (ablation) | ODCS on ResNet-20, ResNet-56, and ResNet-110 · CIFAR-10 | accuracy: 92.28 -> 93.13; 93.45 -> 94.29; 93.01 -> 94.55 | Helped |
| Allow kernel norm to vary freely rather than constraining it to 1 (ablation) | ODCS on ResNet-20 · CIFAR-10 | accuracy: 91.46 -> 93.13 | Helped |
Closest papers by meaning
- An Adaptive Orthogonal Convolution Scheme for Efficient and Flexible CNN Architectures (ICML 2025 · similarity 0.87)
- Orthogonal Convolutional Neural Networks (CVPR 2020 · similarity 0.87)
- Can We Gain More from Orthogonality Regularizations in Training Deep Networks? (NeurIPS 2018 · similarity 0.86)
- Scaling-up Diverse Orthogonal Convolutional Networks by a Paraunitary Framework (ICML 2022 · similarity 0.86)
- Constructing Orthogonal Convolutions in an Explicit Manner (ICLR 2022 · similarity 0.86)
- On the Sparsity-Storage-Accuracy Tradeoff in Parsimoniously Activated Dictionary Learning (NeurIPS 2026 · 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.