Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Bias-Invariant Representation by Cross-Sample Mutual Information Minimization
Wei Zhu, Haitian Zheng, Haofu Liao, Weijian Li, Jiebo Luo
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 cross-sample estimator I_CS with content representation instead of I_JSD (ablation) | CSAD · Colored MNIST (sigma^2=0.020) | accuracy: 0.896 -> 0.933 | Improved |
| add structural representation alone instead of I_JSD (CSAD-Struc vs AD-JSD) (ablation) | CSAD · Colored MNIST (sigma^2=0.020) | accuracy: 0.896 -> 0.928 | Improved |
| use joint content and structural features instead of content only (ablation) | CSAD · Colored MNIST (sigma^2=0.020) | accuracy: 0.933 -> 0.943 | Improved |
| use joint features instead of structural only (ablation) | CSAD · Colored MNIST (sigma^2=0.020) | accuracy: 0.928 -> 0.943 | Improved |
| add cross-sample debiasing under milder bias (sigma^2=0.050) (ablation) | CSAD vs Ruggero et al. · Colored MNIST (sigma^2=0.050) | accuracy: 0.982 -> 0.985 | No clear effect |
| CSAD vs best baseline Ruggero et al. under severe bias | CSAD · Colored MNIST (sigma^2=0.020) | accuracy: 0.864 -> 0.943 | Improved |
| CSAD vs Kim et al. on IMDB face | CSAD · IMDB face (trained on EB2, test) | accuracy: 0.7450 -> 0.7865 | Improved |
| CSAD vs Ruggero et al. on IMDB face trained on EB1, test | CSAD · IMDB face (trained on EB1, test) | accuracy: 0.8720 -> 0.8696 | No clear effect |
| CSAD vs LfF on bias-conflicting set | CSAD · CelebA BlondHair | accuracy: 0.8340 -> 0.8753 | Improved |
| CSAD vs Kim et al. on mPower | CSAD · mPower tapping | AUC: 0.759 -> 0.772 | Improved |
| increase lambda from 1 to 20 (ablation) | CSAD · Adult | balanced accuracy: 81.4 -> 78.9 | Mixed |
| CSAD vs baseline | CSAD · Adult | Gap_G RMS: .179 -> .065 | Improved |
Closest papers by meaning
- BiasAdv: Bias-Adversarial Augmentation for Model Debiasing (CVPR 2023 · similarity 0.88)
- Self-supervised Debiasing Using Low Rank Regularization (CVPR 2024 · similarity 0.86)
- DeNetDM: Debiasing by Network Depth Modulation (NeurIPS 2024 · similarity 0.85)
- A Simple Remedy for Dataset Bias via Self-Influence: A Mislabeled Sample Perspective (NeurIPS 2024 · similarity 0.85)
- Diffusing DeBias: Synthetic Bias Amplification for Model Debiasing (NeurIPS 2025 · similarity 0.85)
- Enhancing Intrinsic Features for Debiasing via Investigating Class-Discerning Common Attributes in Bias-Contrastive Pair (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.
