Research map / Learning theory and algorithms / Fairness
Flexibly Fair Representation Learning by Disentanglement
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 |
|---|---|---|---|
| use attribute-aligned sensitive latent dimensions and remove the relevant dimensions at test time | FFVAE · DSpritesUnfair fair classification with correlated Shape and XPosition | demographic parity distance | Helped |
| use a compositional representation with multiple sensitive dimensions | FFVAE · DSpritesUnfair conjunction subgroups | accuracy at a given demographic parity distance | Helped |
| use sensitive attributes as training-time side information | FFVAE · Communities & Crime | accuracy | Helped |
| use FFVAE instead of β-VAE | Celeb-A · HeavyMakeup fair classification, ¬Eyeglasses ∧ Male subgroup | accuracy at demographic parity distance | Helped |
| use FFVAE instead of β-VAE | Celeb-A · HeavyMakeup fair classification, ¬Chubby ∧ ¬Eyeglasses subgroup | low demographic parity distance solution | Hurt |
| increase the predictiveness weight α (ablation) | FFVAE · DSpritesUnfair predictiveness audit | sensitive-attribute prediction loss | Helped |
| increase the predictiveness weight α (ablation) | FFVAE · DSpritesUnfair disentanglement audit | sensitive-attribute prediction loss from representation with b_i removed | Helped |
| set α = 0, removing the sensitive-attribute predictiveness term (ablation) | FFVAE (FactorVAE variant) · DSpritesUnfair disentanglement audit | sensitive-attribute prediction loss | Hurt |
| increase α (ablation) | FFVAE · DSpritesUnfair mutual information gap | MIG | Helped |
| increase γ at high α (ablation) | FFVAE · DSpritesUnfair mutual information gap | MIG | Helped |
| increase γ at low α (ablation) | FFVAE · DSpritesUnfair mutual information gap | MIG | Hurt |
| set γ = 0, removing the disentanglement prior (ablation) | FFVAE (CVAE variant) · fair classification audits | fairness-accuracy tradeoff | Mixed |
Closest papers by meaning
- On the Fairness of Disentangled Representations (NeurIPS 2019 · similarity 0.91)
- Disentangled Representation Learning with the Gromov-Monge Gap (ICLR 2025 · similarity 0.86)
- FADES: Fair Disentanglement with Sensitive Relevance (CVPR 2024 · similarity 0.86)
- Enriching Disentanglement: From Logical Definitions to Quantitative Metrics (NeurIPS 2024 · similarity 0.85)
- Disentangled Representation Learning via Flow Matching (NeurIPS 2026 · similarity 0.85)
- Disentangling, Amplifying, and Debiasing: Learning Disentangled Representations for Fair Graph Neural Networks (AAAI 2025 · 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.