Fair Normalizing Flows
arXiv:2106.05937
Abstract
Fair representation learning is an attractive approach that promises fairness of downstream predictors by encoding sensitive data. Unfortunately, recent work has shown that strong adversarial predictors can still exhibit unfairness by recovering sensitive attributes from these representations. In this work, we present Fair Normalizing Flows (FNF), a new approach offering more rigorous fairness guarantees for learned representations. Specifically, we consider a practical setting where we can estimate the probability density for sensitive groups. The key idea is to model the encoder as a normalizing flow trained to minimize the statistical distance between the latent representations of different groups. The main advantage of FNF is that its exact likelihood computation allows us to obtain guarantees on the maximum unfairness of any potentially adversarial downstream predictor. We experimentally demonstrate the effectiveness of FNF in enforcing various group fairness notions, as well as other attractive properties such as interpretability and transfer learning, on a variety of challenging real-world datasets.
References in corpus (21)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Auto-Encoding Variational Bayes
- WaveNet: A Generative Model for Raw Audio
- Equality of Opportunity in Supervised Learning
- Pixel Recurrent Neural Networks
- NICE: Non-linear Independent Components Estimation
- Variational Inference with Normalizing Flows
- Density estimation using Real NVP
- Actionable Recourse in Linear Classification
- MADE: Masked Autoencoder for Distribution Estimation
- Anomaly Detection with Density Estimation
- Generative Probabilistic Novelty Detection with Adversarial Autoencoders
- Flexibly Fair Representation Learning by Disentanglement
- Overlearning Reveals Sensitive Attributes
- Learning Controllable Fair Representations
- Learning perturbation sets for robust machine learning
- Interpreting Neural Network Judgments via Minimal, Stable, and Symbolic Corrections
- A Closer Look at the Optimization Landscapes of Generative Adversarial Networks
- A Theory of Usable Information Under Computational Constraints
- Inherent Tradeoffs in Learning Fair Representations
- Learning Fair and Interpretable Representations via Linear Orthogonalization