analiza aiqbench
This research addresses a critical fairness issue in differentially private machine learning where adaptive gradient clipping disproportionately harms minority groups by over-suppressing their gradients. The proposed bounded adaptive clipping method introduces a tunable lower bound to mitigate this bias, achieving significant improvements in worst-class accuracy across standard benchmarks. For practitioners deploying privacy-preserving models, this work offers a practical mechanism to balance privacy guarantees with equitable model performance without sacrificing the core utility of differential privacy.
Mitigating Disparate Impact of Differentially Private Learning through Bounded Adaptive Clipping
arXiv:2506.01396v2 Announce Type: replace Abstract: Differential privacy (DP) has become an essential framework for privacy-preserving machine learning. Existing DP learning methods, however, often have disparate impacts on model predictions, e.g., for minority groups. Gradient clipping, which is often used in DP learning, can suppress larger gradients from challenging samples. We show that this problem is amplified by adaptive clipping, which will often shrink the clipping bound to tiny values to match a well-fitting majority, while significantly reducing the accuracy for others. We propose bounded adaptive clipping, which introduces a tunable lower bound to prevent excessive gradient suppression. Our method improves worst-class accuracy by over 10 percentage points on Skewed and Fashion MNIST compared to unbounded adaptive clipping, 7 points compared to Automatic clipping, and 5 points compared to constant clipping. The code is available at https://github.com/TrustworthyMLHelsinki/adaptive-clipping-fairness.
Abstract:Differential privacy (DP) has become an essential framework for privacy-preserving machine learning. Existing DP learning methods, however, often have disparate impacts on model predictions, e.g., for minority groups. Gradient clipping, which is often used in DP learning, can suppress larger gradients from challenging samples. We show that this problem is amplified by adaptive clipping, which will often shrink the clipping bound to tiny values to match a well-fitting majority, while significantly reducing the accuracy for others. We propose bounded adaptive clipping, which introduces a tunable lower bound to prevent excessive gradient suppression. Our method improves worst-class accuracy by over 10 percentage points on Skewed and Fashion MNIST compared to unbounded adaptive clipping, 7 points compared to Automatic clipping, and 5 points compared to constant clipping. The code is available at this https URL.
Machine Learning (cs.LG); Cryptography and Security (cs.CR); Machine Learning (stat.ML)
arXiv:2506.01396 [cs.LG]
(or arXiv:2506.01396v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2506.01396 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 232 cuvinte
Subiecte conexe
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
Apple Workshop on Privacy-Preserving Machine Learning & AI 2026
Apple · 8 mai 2026
BRICS Draws a New Global Line: Full Text of New Delhi Declaration
slguardian.org · 12 sept. 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · 5 oct. 2026