Learning Adversarially Fair and Transferable Representations (2018-02-17T00:00:00.000000Z)

TL;DR

This paper presents the first in-depth experimental demonstration of fair transfer learning and demonstrates empirically that the authors' learned representations admit fair predictions on new tasks while maintaining utility, an essential goal of fair representation learning.

Abstract

In this paper, we advocate for representation learning as the key to mitigating unfair prediction outcomes downstream. Motivated by a scenario where learned representations are used by third parties with unknown objectives, we propose and explore adversarial representation learning as a natural method of ensuring those parties act fairly. We connect group fairness (demographic parity, equalized odds, and equal opportunity) to different adversarial objectives. Through worst-case theoretical guarantees and experimental validation, we show that the choice of this objective is crucial to fair prediction. Furthermore, we present the first in-depth experimental demonstration of fair transfer learning and demonstrate empirically that our learned representations admit fair predictions on new tasks while maintaining utility, an essential goal of fair representation learning.

Authors

R. Zemel

11 papers

David Madras

1 papers

Elliot Creager

1 papers

TL;DR

Abstract

Authors

References38 items

Mitigating Unwanted Biases with Adversarial Learning

Calibration for the (Computationally-Identifiable) Masses

Predict Responsibly: Increasing Fairness by Learning To Defer

Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness

Provably Fair Representations

On Fairness and Calibration

Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations

Learning Fair Classifiers: A Regularization-Inspired Approach

On fairness

Optimized Data Pre-Processing for Discrimination Prevention

Improved Training of Wasserstein GANs

Semantic Segmentation using Adversarial Networks

Fairness Beyond Disparate Treatment & Disparate Impact: Learning Classification without Disparate Mistreatment

Fair Prediction with Disparate Impact: A Study of Bias in Recidivism Prediction Instruments

Equality of Opportunity in Supervised Learning

Inherent Trade-Offs in the Fair Determination of Risk Scores

Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings

Improved Techniques for Training GANs

Semi-Supervised Learning with Generative Adversarial Networks

Censoring Representations with an Adversary

The Variational Fair Autoencoder

Discrimination- and privacy-aware patterns

Domain-Adversarial Training of Neural Networks

Adam: A Method for Stochastic Optimization

Learning Fair Representations

Fairness-Aware Classifier with Prejudice Remover Regularizer

Representation Learning: A Review and New Perspectives

Fairness through awareness

Noise-contrastive estimation: A new estimation principle for unnormalized statistical models

A Kernel Method for the Two-Sample-Problem

Domain Adaptation with Structural Correspondence Learning

Elements of Information Theory

Learning Factorial Codes by Predictability Minimization

GENERATIVE ADVERSARIAL NETS

Optimized Pre-Processing for Discrimination Prevention

Semantic Segmentation with

Reducing the Dimensionality of Data with Neural

Supporting Online Material for Reducing the Dimensionality of Data with Neural Networks

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names