Fairness under Covariate Shift: Improving Fairness-Accuracy tradeoff with few Unlabeled Test Samples
Not on the map. Why?
Abstract (arXiv)
Covariate shift in the test data is a common practical phenomena that can significantly downgrade both the accuracy and the fairness performance of the model. Ensuring fairness across different sensitive groups under covariate shift is of paramount importance due to societal implications like criminal justice. We operate in the unsupervised regime where only a small set of unlabeled test samples along with a labeled training set is available. Towards improving fairness under this highly challenging yet realistic scenario, we make three contributions. First is a novel composite weighted entropy based objective for prediction accuracy which is optimized along with a representation matching loss for fairness. We experimentally verify that optimizing with our loss formulation outperforms a number of state-of-the-art baselines in the pareto sense with respect to the fairness-accuracy tradeoff on several standard datasets. Our second contribution is a new setting we term Asymmetric Covariate Shift that, to the best of our knowledge, has not been studied before. Asymmetric covariate shift occurs when distribution of covariates of one group shifts significantly compared to the other groups and this happens when a dominant group is over-represented. While this setting is extremely challenging for current baselines, We show that our proposed method significantly outperforms them. Our third contribution is theoretical, where we show that our weighted entropy term along with prediction loss on the training set approximates test loss under covariate shift. Empirically and through formal sample complexity bounds, we show that this approximation to the unseen test loss does not depend on importance sampling variance which affects many other baselines.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 41 of its 67 references that are in the map. First posted on arXiv 11 Oct 2023.
Builds on
41 of 67 references.
References · 41
- 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 4,993
- 2016Equality of Opportunity in Supervised Learningcited by 2,000
- 2014Unsupervised Domain Adaptation by Backpropagationcited by 1,906
- 2019Invariant Risk Minimizationcited by 1,832
- 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
- 2019Self-training with Noisy Student improves ImageNet classificationcited by 1,304
- 2019Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationcited by 1,232
- 2014Certifying and removing disparate impactcited by 715
- 2018Mitigating Unwanted Biases with Adversarial Learningcited by 663
- 2018A Reductions Approach to Fair Classificationcited by 585
- 2020Improving robustness against common corruptions by covariate shift adaptationcited by 427
- 2017On Fairness and Calibrationcited by 406
29 more
- 2017Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairnesscited by 381
- 2016Fairness Beyond Disparate Treatment & Disparate Impact: Learning Classification without Disparate Mistreatmentcited by 368
- 2018A Survey of Unsupervised Deep Domain Adaptationcited by 361
- 2018Learning Adversarially Fair and Transferable Representationscited by 353
- 2018Learning Models with Uniform Performance via Distributionally Robust Optimizationcited by 334
- 2016Revisiting Batch Normalization For Practical Domain Adaptationcited by 285
- 2021Gradient Matching for Domain Generalizationcited by 258
- 2020Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shiftcited by 235
- 2018Empirical Risk Minimization under Fairness Constraintscited by 226
- 2020Fairness in Machine Learningcited by 194
- 2021A Fine-Grained Analysis on Distribution Shiftcited by 171
- 2018Classification with Fairness Constraints: A Meta-Algorithm with Provable Guaranteescited by 146
- 2019Inherent Tradeoffs in Learning Fair Representationscited by 144
- 2019Wasserstein Fair Classificationcited by 131
- 2020A survey on domain adaptation theory: learning bounds and theoretical guaranteescited by 109
- 2018Optimization with Non-Differentiable Constraints with Applications to Fairness, Recall, Churn, and Other Goalscited by 103
- 2020Fair Regression with Wasserstein Barycenterscited by 99
- 2016A Confidence-Based Approach for Balancing Fairness and Accuracycited by 73
- 2020Opportunities and Challenges in Deep Learning Adversarial Robustness: A Surveycited by 71
- 2020Robust Fairness under Covariate Shiftcited by 58
- 2021FARF: A Fair and Adaptive Random Forests Classifiercited by 37
- 2020Ensuring Fairness Beyond the Training Datacited by 35
- 2019Fairness Warnings and Fair-MAML: Learning Fairly with Minimal Datacited by 23
- 2022Transferring Fairness under Distribution Shifts via Fair Consistency Regularizationcited by 21
- 2019Robust Importance Weighting for Covariate Shiftcited by 14
- 2021Wasserstein Robust Classification with Fairness Constraintscited by 10
- 2021Costs and Benefits of Fair Regressioncited by 8
- 2023Stochastic Re-weighted Gradient Descent via Distributionally Robust Optimizationcited by 6
- 2017Provable benefits of representation learningcited by 3
Built on by
Citing papers per year: 2024: 1; 2025: 1.
2 papers.