The Variational Fair Autoencoder

Abstract

We investigate the problem of learning representations that are invariant to certain nuisance or sensitive factors of variation in the data while retaining as much of the remaining information as possible. Our model is based on a variational autoencoding architecture with priors that encourage independence between sensitive and latent factors of variation. Any subsequent processing, such as classification, can then be performed on this purged latent representation. To remove any remaining dependencies we incorporate an additional penalty term based on the "Maximum Mean Discrepancy" (MMD) measure. We discuss how these architectures can be efficiently trained on data and show in experiments that this method is more effective than previous work in removing unwanted sources of variation while maintaining informative latent representations.

Fixed typo in eq. 3 and 4

Cited by in corpus (36)

• An Introduction to Variational Autoencoders
• Counterfactual Fairness
• Asymmetric Tri-training for Unsupervised Domain Adaptation
• A Clarification of the Nuances in the Fairness Metrics Landscape
• Fairness in Credit Scoring: Assessment, Implementation and Profit Implications
• Learning Representations for Counterfactual Inference
• An Empirical Characterization of Fair Machine Learning For Clinical Risk Prediction
• EDITS: Modeling and Mitigating Data Bias for Graph Neural Networks
• Learning to Protect Communications with Adversarial Neural Cryptography
• Lifelong Generative Modeling
• Unsupervised Domain Adaptation with Adversarial Residual Transform Networks
• Robust Unsupervised Domain Adaptation for Neural Networks via Moment Alignment
• DIVA: Domain Invariant Variational Autoencoders
• Fairness in Algorithmic Decision Making: An Excursion Through the Lens of Causality
• GeneGAN: Learning Object Transfiguration and Attribute Subspace from Unpaired Data
• Towards Fair Classifiers Without Sensitive Attributes: Exploring Biases in Related Features
• Noise-tolerant fair classification
• Fair Representation: Guaranteeing Approximate Multiple Group Fairness for Unknown Tasks
• Disentangled Representation Learning for Astronomical Chemical Tagging
• Adversarial training approach for local data debiasing
• Kernel Two-Sample Tests in High Dimension: Interplay Between Moment Discrepancy and Dimension-and-Sample Orders
• BeFair: Addressing Fairness in the Banking Sector
• Conditional out-of-sample generation for unpaired data using trVAE
• Representation via Representations: Domain Generalization via Adversarially Learned Invariant Representations
• Ethical Adversaries: Towards Mitigating Unfairness with Adversarial Machine Learning
• Benign Shortcut for Debiasing: Fair Visual Recognition via Intervention with Shortcut Features
• Privacy Enhancing Machine Learning via Removal of Unwanted Dependencies
• CUDA: Contradistinguisher for Unsupervised Domain Adaptation
• Learning Smooth and Fair Representations
• Learning Fair and Interpretable Representations via Linear Orthogonalization
• Deep Discriminative Learning for Unsupervised Domain Adaptation
• Variation Network: Learning High-level Attributes for Controlled Input Manipulation
• Fairness constraints can help exact inference in structured prediction
• Scrutinizing and De-Biasing Intuitive Physics with Neural Stethoscopes
• From abstract items to latent spaces to observed data and back: Compositional Variational Auto-Encoder
• Invariant Representations with Stochastically Quantized Neural Networks