Research map / Label-efficient and robust learning / Recognition and long-tail
RBF-Softmax: Learning Deep Representative Prototypes with Radial Basis Function Softmax
Xiao Zhang, Rui Zhao, Yu Qiao, Hongsheng Li
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 |
|---|---|---|---|
| replace softmax inner-product logits with RBF-Softmax | Softmax · ImageNet ResNet-50 | top-1 accuracy: 76.8% -> 77.1% | Improved |
| replace softmax with RBF-Softmax | Softmax · ImageNet EfficientNet-B1 | top-1 accuracy: 75.9% -> 76.6% | Improved |
| replace softmax with RBF-Softmax | Softmax · ImageNet EfficientNet-B4 | top-1 accuracy: 78.8% -> 79.0% | Improved |
| replace softmax with RBF-Softmax | Softmax · CIFAR-10 ResNet-20 | accuracy: 91.25% -> 92.77% | Improved |
| replace softmax with RBF-Softmax | Softmax · MNIST 32-D | accuracy: 99.27% -> 99.73% | Improved |
| replace softmax with RBF-Softmax | Softmax · CIFAR-100 VGGNet-19 | accuracy: 72.23% -> 73.98% | Improved |
| replace softmax with RBF-Softmax | Softmax · CIFAR-100 prototypes | Calinski-Harabaz index: 2.81 ± 0.21 -> 7.33 ± 0.25 | Improved |
| replace softmax with RBF-Softmax | Softmax · CIFAR-100 prototypes | similarity to WordNet matrix: 0.17 ± 0.05 -> 0.36 ± 0.11 | Improved |
| increase gamma from 1.0 to 2.0 (ablation) | RBF-Softmax · MNIST 32-D | accuracy: 99.71% -> 99.42% | Hurt |
| increase gamma from 1.0 to 4.0 (s=12) (ablation) | RBF-Softmax · CIFAR-100 VGGNet-19 | accuracy: 72.62%±0.05% -> 71.77%±0.04% | Hurt |
| increase s from 2.0 to 8.0 (ablation) | RBF-Softmax · MNIST 32-D | accuracy: 99.71% -> 99.61% | Hurt |
| increase s from 10 to 14 (gamma=2.2) (ablation) | RBF-Softmax · CIFAR-100 VGGNet-19 | accuracy: 72.72%±0.03% -> 73.98%±0.02% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Large-Margin Softmax Loss for Convolutional Neural Networks (ICML 2016 · similarity 0.89)
- Learning Towards The Largest Margins (ICLR 2022 · similarity 0.89)
- Killing Two Birds With One Stone: Efficient and Robust Training of Face Recognition CNNs by Partial FC (CVPR 2022 · similarity 0.87)
- Back to the Future: Radial Basis Function Network Revisited (TPAMI 2019 · similarity 0.87)
- Dataset Distillation with Neural Characteristic Function: A Minmax Perspective (CVPR 2025 · similarity 0.85)
- MaxQ: Multi-Axis Query for N:M Sparsity Network (CVPR 2024 · similarity 0.85)
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.
