Optimal Cox regression under federated differential privacy: coefficients and cumulative hazards
Elly K. H. Hung, Yi Yu · 2025
Read Optimal Cox regression under federated differential privacy: coefficients and cumulative hazards by Elly K. H. Hung, Yi Yu (2025) 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-2508-19640.txt.
Description
Elly Hung and Yi Yu set the price of privacy for survival analysis across institutions in a 2025 stats paper. Hospitals holding separate patient cohorts want joint Cox regression, a standard survival model, without pooling records. Federated change-based privacy formalizes the protection each site must give. The authors tackle two targets, regression coefficients and cumulative hazard curves, while sites differ in sample size and privacy budget.
Matching lower and upper bounds, tight up to logarithmic factors, reveal sharp phase transitions where privacy starts to cost accuracy. Fast methods adapt private binary trees and gradient descent to the model's dependent estimating math rules. Companion results cover usual test hazards, survival curves, and a relaxed label-only privacy notion, with software put out.
Catalog record
| CREATOR | Elly K. H. Hung, Yi Yu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 163 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-19640