Research map / Adversarial robustness and security / Certified and empirical robustness
Group-Wise Inhibition Based Feature Regularization for Robust Classification
Haozhe Liu, Haoqian Wu, Weicheng Xie, Feng Liu, Linlin Shen
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 channel-wise inhibition to the baseline (ablation) | ResNet-18 · PASCAL VOC 2012 val (ImageNet-pretrained) | mean AP: 77.1 -> 81.2 | Improved |
| use group-wise inhibition (CFG + GMW) instead of baseline (ablation) | ResNet-18 · PASCAL VOC 2012 val (ImageNet-pretrained) | mean AP: 77.1 -> 81.8 | Improved |
| add orthogonal loss L_o to group-wise inhibition (ablation) | TENET · PASCAL VOC 2012 val (ImageNet-pretrained) | mean AP: 81.8 -> 82.3 | Improved |
| replace binary mask with smoothed reversed maps (RRF) (ablation) | TENET · PASCAL VOC 2012 val (ImageNet-pretrained) | mean AP: 81.1 -> 82.3 | Improved |
| replace instance-wise Grad-CAM inhibition with group-wise CFG/GMW (ablation) | TENET · PASCAL VOC 2012 val (ImageNet-pretrained) | mean AP: 81.3 -> 82.3 | Improved |
| add TENET regularization (vs GoCNN, no annotations) | ResNet-18 · PASCAL VOC 2012 val (ImageNet-pretrained) | mean AP: 79.4 -> 82.3 | Improved |
| combine TENET with Augmix | ResNet-50 · ImageNet-C (90-epoch protocol) | mCE: 75.3 -> 71.1 | Improved |
| TENET (Aug.) vs Augmix | ResNet-50 · ImageNet-C (180-epoch protocol) | mCE: 70.5 -> 69.6 | Improved |
| add TENET to Augmix | ResNeXt-29 · CIFAR-10-C | mCE: 13.32 -> 12.31 | Improved |
| add TENET to Augmix | ResNeXt-29 · CIFAR-10 FGSM (8/255) | top-1 error: 76.03 -> 60.47 | Improved |
| add TENET to adversarial training | ResNeXt-29 · CIFAR-10 FGSM (8/255) | top-1 error: 36.37 -> 31.75 | Improved |
| add TENET to adversarial training | ResNeXt-29 · CIFAR-100 PGD-100 (8/255) | top-1 error: 65.17 -> 63.13 | Improved |
Closest papers by meaning
- Group-Wise Dynamic Dropout Based on Latent Semantic Variations (AAAI 2020 · similarity 0.86)
- Robust Learning with Progressive Data Expansion Against Spurious Correlation (NeurIPS 2023 · similarity 0.86)
- Group Downsampling with Equivariant Anti-aliasing (ICLR 2025 · similarity 0.86)
- GSE: Group-wise Sparse and Explainable Adversarial Attacks (ICLR 2025 · similarity 0.85)
- On the VC dimension of deep group convolutional neural networks (NeurIPS 2025 · similarity 0.85)
- Label-Efficient Group Robustness via Out-of-Distribution Concept Curation (CVPR 2024 · 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.
