The Privacy Price of Tail-Risk Learning: Effective Tail Sample Size in Differentially Private CVaR Optimization
El Mustapha Mansouri · 2026
Read The Privacy Price of Tail-Risk Learning: Effective Tail Sample Size in Differentially Private CVaR Optimization by El Mustapha Mansouri (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-2605-16219.txt.
Description
Mansouri prices the privacy cost of tail-risk learning under differential privacy. For tail mass tau, the privacy sample size is not n but n times tau, so the effective private tail size is epsilon times n times tau. Private CVaR excess risk splits into ordinary tail error plus a privacy price. Complete rates cover scalar estimation and finite classes under pure differential privacy, with lower bounds extending to approximate privacy in small-delta regimes. The canonical hard case is ordinary private learning on the tail records.
Catalog record
| CREATOR | El Mustapha Mansouri |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 78204 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.16219