Differentially Private Non-convex Distributionally Robust Optimization
Difei Xu, Meng Ding, Zebin Ma, Huanyi Xie, Youming Tao, Aicha Slaitane, Di Wang · 2026
Read Differentially Private Non-convex Distributionally Robust Optimization by Difei Xu, Meng Ding, Zebin Ma, Huanyi Xie, Youming Tao, Aicha Slaitane, Di Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-16155.txt.
Description
Difei Xu leads a team of seven researchers bringing differential privacy to distributionally robust optimization with non-convex losses. Real deployments face distribution shifts, group imbalances, and adversarial perturbations. These can badly degrade standard empirical risk minimization. Robust optimization counters this by optimizing worst-case expected loss over an uncertainty set of distributions. Yet training data there is sensitive and needs privacy protection. That combination has had far less study than private standard training.
The team studies private finite-sum robust optimization under general divergence measures. They recast it as a minimization problem and develop a method called DP Double-Spider. Under mild assumptions it reaches a stated utility bound in gradient norm. For KL-divergence, a second method, DP Recursive-Spider, matches the best-known rate for non-convex private standard training. Experiments show both methods beating existing approaches for private minimax optimization.
Catalog record
| CREATOR | Difei Xu, Meng Ding, Zebin Ma, Huanyi Xie, Youming Tao, Aicha Slaitane, Di Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 89 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-16155