Research map / Label-efficient and robust learning / Self-supervised and contrastive
Contrastive Learning with the Feature Reconstruction Amplifier
Wentao Cui, Liang Bai
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 FRA with Student-t loss | SimCLR · CIFAR-10 linear evaluation | accuracy: 89.16 -> 90.72 | Improved |
| Add FRA with Student-t loss | SimCLR · CIFAR-100 linear evaluation | accuracy: 64.23 -> 66.72 | Improved |
| Add FRA with Student-t loss | SimCLR · STL-10 linear evaluation | accuracy: 87.40 -> 91.07 | Improved |
| Add FRA with Student-t loss | SimCLR · Voc2007 linear evaluation | mAP: 61.13 -> 62.61 | Improved |
| Add FRA with Student-t loss | SimCLR · ImageNet-100 linear evaluation | Top-1 accuracy: 69.03 -> 69.84 | Improved |
| Add FRA with Student-t loss | SimCLR · ImageNet-100 linear evaluation | Top-5 accuracy: 90.21 -> 90.60 | Improved |
| Add FRA with Student-t loss | SimCLR · Transfer across CIFAR-10, CIFAR-100, and STL-10 | accuracy: 59.81 -> 63.86; 72.59 -> 75.14; 83.16 -> 84.99; 70.21 -> 72.58; 84.15 -> 86.78; 57.24 -> 61.68 | Improved |
| Add FRA using MSE loss (ablation) | SimCLR · CIFAR-10 linear evaluation | accuracy: 87.98 -> 88.56 | Improved |
| Add FRA using InfoNCE loss (ablation) | SimCLR · CIFAR-10 linear evaluation | accuracy: 87.98 -> 88.31 | Improved |
| Add FRA using Student-t loss (ablation) | SimCLR · CIFAR-10 linear evaluation | accuracy: 87.98 -> 89.27 | Improved |
| Use a single-BN-layer amplifier head (ablation) | SimFRA · STL-10 and ImageNet-100 training | training stability | Hurt |
Closest papers by meaning
- Learning Contrastive Embedding in Low-Dimensional Space (NeurIPS 2022 · similarity 0.90)
- Improving Contrastive Learning by Visualizing Feature Transformation (ICCV 2021 · similarity 0.88)
- A Broad Study on the Transferability of Visual Representations With Contrastive Learning (ICCV 2021 · similarity 0.88)
- The Loss Is Not Enough: Sampling Conditions and Inductive Bias in Contrastive Representation Learning (ICML 2026 · similarity 0.87)
- ArCL: Enhancing Contrastive Learning with Augmentation-Robust Representations (ICLR 2023 · similarity 0.87)
- The Power of Contrast for Feature Learning: A Theoretical Analysis (JMLR 2023 · similarity 0.87)
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.
