Research map / Label-efficient and robust learning / Self-supervised and contrastive
ProtoFair: Fair Self-Supervised Contrastive Learning via Pseudo-Counterfactual Pairs
Marah Halawa, Olaf Hellwich
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 ProtoFair regularizer | SupCon · CelebA T:b,S:y | EO: 16.9 -> 6.9 | Improved |
| add ProtoFair regularizer | SupCon · CelebA T:a,S:m | EO: 30.5 -> 14.2 | Improved |
| add ProtoFair regularizer | SupCon · CelebA T:e,S:y | EO: 10.8 -> 1.9 | Improved |
| add ProtoFair regularizer | SupCon · CelebA T:a,S:y | EO: 21.7 -> 17.7 | Improved |
| add ProtoFair regularizer | SimCLR · CelebA T:a,S:m (no target labels) | EO: 29.4 -> 21.9, ACC 75.7 -> 73.3 | Improved |
| add ProtoFair regularizer | BarlowTwins · CelebA T:b,S:m (no target labels) | EO: 12.07 -> 5.98 | Improved |
| add ProtoFair regularizer | BYOL · CelebA T:e,S:m (no target labels) | EO: 17.04 -> 7.11 | Improved |
| add ProtoFair regularizer | SupCon · UTKFace alpha=4 | EO: 10.6 -> 6.6 | Improved |
| add ProtoFair regularizer | SupCon · NIH Chest X-rays | AUROC gap: 0.03 -> 0.01, AUROC 0.70 -> 0.72 | Improved |
| add ProtoFair regularizer | SupCon · CelebA Bags Under Eyes sensitive probe | probe accuracy: 82.41 -> 76.07 | Improved |
| increase fairness weight lambda to 0.7 or higher (ablation) | ProtoFair · CelebA (four T,S pairs) | accuracy | Hurt |
| vary number of prototypes K (ablation) | ProtoFair · CelebA (four T,S pairs) | accuracy: accuracy within 83.0--84.5 | No clear effect |
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.
- BarlowTwins — Barlow Twins: Self-Supervised Learning via Redundancy Reduction (ICML 2021) · better on 2 of 2 shared results
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
- Fairness-aware Contrastive Learning with Partially Annotated Sensitive Attributes (ICLR 2023 · similarity 0.86)
- On the duality between contrastive and non-contrastive self-supervised learning (ICLR 2023 · similarity 0.85)
- Solving Inefficiency of Self-Supervised Representation Learning (ICCV 2021 · similarity 0.85)
- Divide and Contrast: Self-Supervised Learning From Uncurated Data (ICCV 2021 · similarity 0.85)
- Collapse-Proof Non-Contrastive Self-Supervised Learning (ICML 2025 · similarity 0.85)
- Do Counterfactually Fair Image Classifiers Satisfy Group Fairness? -- A Theoretical and Empirical Study (NeurIPS 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.
