Flexibly Fair Representation Learning by Disentanglement

Abstract

We consider the problem of learning representations that achieve group and subgroup fairness with respect to multiple sensitive attributes. Taking inspiration from the disentangled representation learning literature, we propose an algorithm for learning compact representations of datasets that are useful for reconstruction and prediction, but are also \emph{flexibly fair}, meaning they can be easily modified at test time to achieve subgroup demographic parity with respect to multiple sensitive attributes and their conjunctions. We show empirically that the resulting encoder---which does not require the sensitive attributes for inference---enables the adaptation of a single representation to a variety of fair classification tasks with new target labels and subgroup definitions.

Cited by in corpus (46)

• Shortcut Learning in Deep Neural Networks
• A survey of recent methods for addressing AI fairness and bias in biomedicine
• Weakly-Supervised Disentanglement Without Compromises
• An analysis on the use of autoencoders for representation learning: fundamentals, learning task case studies, explainability and challenges
• On Disentangled Representations Learned From Correlated Data
• Leakage of Dataset Properties in Multi-Party Machine Learning
• Learning Certified Individually Fair Representations
• Reducing Sentiment Bias in Language Models via Counterfactual Evaluation
• Improving the Fairness of Deep Generative Models without Retraining
• DeBayes: a Bayesian Method for Debiasing Network Embeddings
• Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigation
• FairFaceGAN: Fairness-aware Facial Image-to-Image Translation
• Deformable Generator Networks: Unsupervised Disentanglement of Appearance and Geometry
• Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding
• Conditional Learning of Fair Representations
• Auditing and Achieving Intersectional Fairness in Classification Problems
• Adversarial training approach for local data debiasing
• Tuning Fairness by Balancing Target Labels
• Say No to the Discrimination: Learning Fair Graph Neural Networks with Limited Sensitive Attribute Information
• Guided Variational Autoencoder for Disentanglement Learning
• README: REpresentation learning by fairness-Aware Disentangling MEthod
• Disentangled Representation with Causal Constraints for Counterfactual Fairness
• A Bandit-Based Algorithm for Fairness-Aware Hyperparameter Optimization
• Benchmarks, Algorithms, and Metrics for Hierarchical Disentanglement
• Impossibility results for fair representations
• An Image is Worth More Than a Thousand Words: Towards Disentanglement in the Wild
• Addressing the Topological Defects of Disentanglement via Distributed Operators
• VarFA: A Variational Factor Analysis Framework For Efficient Bayesian Learning Analytics
• FairMask: Better Fairness via Model-based Rebalancing of Protected Attributes
• Learning Disentangled Representations with Latent Variation Predictability
• Learning Smooth and Fair Representations
• An Attribute-Aligned Strategy for Learning Speech Representation
• Removing Spurious Features can Hurt Accuracy and Affect Groups Disproportionately
• Fair Normalizing Flows
• Generating private data with user customization
• InfoScrub: Towards Attribute Privacy by Targeted Obfuscation
• Fairness in Deep Learning: A Computational Perspective
• A Commentary on the Unsupervised Learning of Disentangled Representations
• Fairness-aware Outlier Ensemble
• Projection-wise Disentangling for Fair and Interpretable Representation Learning: Application to 3D Facial Shape Analysis
• Deep Poisoning: Towards Robust Image Data Sharing against Visual Disclosure
• Disentangling Representations of Text by Masking Transformers
• Boxhead: A Dataset for Learning Hierarchical Representations
• Individually Fair Gradient Boosting
• Group-disentangled Representation Learning with Weakly-Supervised Regularization
• A Sociotechnical View of Algorithmic Fairness