Research map / Learning theory and algorithms / Fairness
FairBatch: Batch Selection for Model Fairness
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 uniform minibatch sampling with adaptive FairBatch group ratios | Logistic regression · Synthetic | EO disparity: .115±.000 -> .012±.001 | Helped |
| Replace uniform minibatch sampling with adaptive FairBatch group ratios | Logistic regression · COMPAS | EO disparity: .239±.006 -> .022±.005 | Helped |
| Replace uniform minibatch sampling with adaptive FairBatch group ratios | Logistic regression · AdultCensus | EO disparity: .054±.005 -> .011±.003 | Helped |
| Use adaptive batch selection for demographic parity | Logistic regression · Synthetic | DP disparity: .257±.000 -> .040±.001 | Helped |
| Use adaptive batch selection for demographic parity | Logistic regression · COMPAS | DP disparity: .192±.006 -> .036±.023 | Helped |
| Use adaptive batch selection for demographic parity | Logistic regression · AdultCensus | DP disparity: .125±.001 -> .010±.005 | Helped |
| Use adaptive batch selection for equalized odds | Logistic regression · Synthetic | ED disparity: .115±.000 -> .038±.002 | Helped |
| Use adaptive batch selection for equalized odds | Logistic regression · COMPAS | ED disparity: .239±.006 -> .052±.014 | Helped |
| Use adaptive batch selection for equalized odds | Logistic regression · AdultCensus | ED disparity: .056±.003 -> .036±.002 | Helped |
| Fine-tune a pretrained model using FairBatch | ResNet18 · UTKFace (gender label) | ED disparity: .086±.012 -> .063±.013 | Helped |
| Fine-tune a pretrained model using FairBatch | GoogLeNet · UTKFace (gender label) | ED disparity: .105±.016 -> .061±.006 | Helped |
| Combine FairBatch with loss-based importance weighting | FairBatch on synthetic data · Synthetic | epochs to convergence: about 50 fewer epochs | Helped |
Closest papers by meaning
- Fairness without Harm: An Influence-Guided Active Sampling Approach (NeurIPS 2024 · similarity 0.88)
- Fair Bilevel Neural Network (FairBiNN): On Balancing fairness and accuracy via Stackelberg Equilibrium (NeurIPS 2024 · similarity 0.88)
- FairNet: Dynamic Fairness Correction without Performance Loss via Contrastive Conditional LoRA (NeurIPS 2025 · similarity 0.88)
- FairWASP: Fast and Optimal Fair Wasserstein Pre-processing (AAAI 2024 · similarity 0.87)
- Navigating Towards Fairness with Data Selection (AAAI 2025 · similarity 0.87)
- Achievable Fairness on Your Data With Utility Guarantees (NeurIPS 2024 · 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.