Research map / Learning theory and algorithms / Fairness
Certification of Distributional Individual Fairness
Matthew Robert Wicker, Vihari Piratla, Adrian Weller
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 |
|---|---|---|---|
| train with F-IBP local IF regularization instead of FTU | 2-layer MLP · Folktables Income | A-DFC: 1.000 -> 0.130 | Improved |
| train with F-IBP local IF regularization instead of FTU | 2-layer MLP · Folktables Income | accuracy: 0.820 -> 0.771 | Hurt |
| train with SenSR instead of FTU | 2-layer MLP · Folktables Employ | A-DFC: 0.999 -> 0.639 | Improved |
| train with SenSR instead of FTU | 2-layer MLP · Folktables Employ | accuracy: 0.809 -> 0.743 | Hurt |
| add L-DIF distributional regularization on top of F-IBP | 2-layer MLP · Folktables Coverage | A-DFC: 0.288 -> 0.136 | Improved |
| add L-DIF distributional regularization on top of F-IBP | 2-layer MLP · Folktables Coverage | accuracy: 0.699 -> 0.698 | No clear effect |
| use U-DIF upper-bound regularization instead of L-DIF | 2-layer MLP · Folktables Income | A-DFC: 0.095 -> 0.042 | Improved |
| use U-DIF upper-bound regularization instead of L-DIF | 2-layer MLP · Folktables Income | accuracy: 0.763 -> 0.725 | Hurt |
| use U-DIF instead of L-DIF | 2-layer MLP · Folktables Employ | E-DFC: 0.051 -> 0.006 | Improved |
| increase network size (number of neurons) | L-DIF certification · DIF upper bound | upper bound value | Hurt |
| increase gamma from 0.01 to 0.2 | DIF upper/lower bounds · 2x16 NN | bound gap: 0.01 -> 0.2 | Hurt |
| train with L-DIF instead of MILP training | single hidden layer NN · training time | training time: 4096 neurons under one hour | Improved |
Closest papers by meaning
- FairProof : Confidential and Certifiable Fairness for Neural Networks (ICML 2024 · similarity 0.88)
- A Comprehensive View of Fairness through Distributional Stability (NeurIPS 2026 · similarity 0.87)
- Certified but Private: Scalable Zero-Knowledge Proofs for the Formal Verification of Neural Networks (NeurIPS 2026 · similarity 0.86)
- Monotone Individual Fairness (ICML 2024 · similarity 0.86)
- Certification-Enhanced Generalization Bounds (NeurIPS 2026 · similarity 0.86)
- On the Maximal Local Disparity of Fairness-Aware Classifiers (ICML 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.
