On the Fairness of Disentangled Representations
arXiv:1905.13662
Abstract
Recently there has been a significant interest in learning disentangled representations, as they promise increased interpretability, generalization to unseen scenarios and faster learning on downstream tasks. In this paper, we investigate the usefulness of different notions of disentanglement for improving the fairness of downstream prediction tasks based on representations. We consider the setting where the goal is to predict a target variable based on the learned representation of high-dimensional observations (such as images) that depend on both the target variable and an \emph{unobserved} sensitive variable. We show that in this setting both the optimal and empirical predictions can be unfair, even if the target variable and the sensitive variable are independent. Analyzing the representations of more than \num{12600} trained state-of-the-art disentangled models, we observe that several disentanglement scores are consistently correlated with increased fairness, suggesting that disentanglement may be a useful property to encourage fairness when sensitive variables are not observed.
References in corpus (42)
- Scikit-learn: Machine Learning in Python
- Auto-Encoding Variational Bayes
- Equality of Opportunity in Supervised Learning
- Counterfactual Fairness
- Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
- Deep Convolutional Inverse Graphics Network
- Disentangling by Factorising
- Adversarially Learned Inference
- Recent Advances in Autoencoder-Based Representation Learning
- Deep Variational Information Bottleneck
- Visual Reinforcement Learning with Imagined Goals
- Towards a Definition of Disentangled Representations
- Weakly-supervised Disentangling with Recurrent Transformations for 3D View Synthesis
- Disentangling factors of variation in deep representations using adversarial training
- Learning Adversarially Fair and Transferable Representations
- Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICA
- On Variational Bounds of Mutual Information
- Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observations
- Discovering Hidden Factors of Variation in Deep Networks
- DARLA: Improving Zero-Shot Transfer in Reinforcement Learning
- A Disentangled Recognition and Nonlinear Dynamics Model for Unsupervised Learning
- Learning Deep Disentangled Embeddings with the F-Statistic Loss
- Are Disentangled Representations Helpful for Abstract Visual Reasoning?
- Structured Disentangled Representations
- Censoring Representations with an Adversary
- On the relation between accuracy and fairness in binary classification
- Disentangled Sequential Autoencoder
- SCAN: Learning Hierarchical Compositional Visual Concepts
- Learning Latent Subspaces in Variational Autoencoders
- Curiosity Driven Exploration of Learned Disentangled Goal Spaces
- Learning Controllable Fair Representations
- On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset
- Bayesian representation learning with oracle constraints
- Disentangling Disentanglement in Variational Autoencoders
- Learning the Irreducible Representations of Commutative Lie Groups
- Disentangling the independently controllable factors of variation by interacting with the world
- Improving Generalization for Abstract Reasoning Tasks Using Disentangled Feature Representations
- Provably Fair Representations
- Understanding Visual Concepts with Continuation Learning
- Learning to Linearize Under Uncertainty
- Learning Disentangled Representations with Reference-Based Variational Autoencoders
- Competitive Training of Mixtures of Independent Deep Generative Models