Research map / Architectures and efficiency / CNN design
SORT: Second-Order Response Transform for Visual Recognition
Yan Wang, Lingxi Xie, Chenxi Liu, Siyuan Qiao, Ya Zhang, Wenjun Zhang, Qi Tian, Alan Yuille
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 element-wise product term to branch sum (ablation) | LeNet · CIFAR-10 | error rate (%): 11.10 -> 10.34 | Improved |
| add element-wise product term to branch sum (ablation) | ResNet-20 · CIFAR-10 | error rate (%): 7.60 -> 7.14 | Improved |
| add element-wise product term to branch sum (ablation) | BigNet · CIFAR-10 | error rate (%): 6.86 -> 6.60 | Improved |
| use product term alone without first-order sum (ablation) | BigNet and ResNet · CIFAR-10 | convergence: non-convergence (-) | Hurt |
| add max on top of sum plus product (ablation) | ResNet-20 · CIFAR-10 | error rate (%): 7.14 -> 7.90 | Hurt |
| add SORT to two-branch network | ResNet-56 · SVHN | error rate (%): 2.49 -> 2.19 | Improved |
| add SORT to two-branch network | ResNet-20 · SVHN | error rate (%): 2.30 -> 2.26 | Improved |
| add SORT to WRN-28 | WRN-28 · SVHN | error rate (%): 1.93 -> 1.48 | Improved |
| add SORT to ResNet-56 | ResNet-56 · CIFAR-10 | error rate (%): 6.30 -> 5.50 | Improved |
| add SORT to ResNet-20 | ResNet-20 · CIFAR-100 | error rate (%): 31.80 -> 31.65 | No clear effect |
| add SORT to ResNet-18 | ResNet-18 · ILSVRC2012 | top-1 error (%): 34.50 -> 32.37 | Improved |
| add SORT to AlexNet* | AlexNet* · ILSVRC2012 | top-1 error (%): 36.71 -> 35.99 | Improved |
Closest papers by meaning
- Is Second-Order Information Helpful for Large-Scale Visual Recognition? (ICCV 2017 · similarity 0.86)
- Statistically-motivated Second-order Pooling (ECCV 2018 · similarity 0.85)
- Second-Order Complexity Theory for Neural Networks (NeurIPS 2026 · similarity 0.85)
- FASON: First and Second Order Information Fusion Network for Texture Recognition (CVPR 2017 · similarity 0.85)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · similarity 0.85)
- REOrdering Patches Improves Vision Models (NeurIPS 2025 · 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.
