Learning with Differentially Private Sliced Wasserstein Gradients

D. Rodriguez, C. Lalanne, J. Loubes

In this work, we introduce a novel framework for privately optimizing objectives that depend on sliced Wasserstein distances between data-dependent empirical measures. Our main theoretical contribution is a non-trivial analysis of the sensitivity of the Wasserstein gradients to individual data points, derived from an explicit formulation of the gradient in a fully discrete setting. This enables strong privacy guarantees with minimal utility loss. We demonstrate that standard privacy accounting methods naturally extend to Wasserstein-based objectives, allowing for large-scale private training. This supports a wide range of private machine learning applications involving distribution matching under privacy constraints on the source, the target, or both. These include: (i) an in-processing method for fairness mitigation using a private Wasserstein penalty, and (ii) what we believe is the first approach for training private sliced Wasserstein autoencoders.

Palabras clave: optimal transport, sliced Wasserstein distance, differential privacy

Machine Learning
2 de septiembre de 2026  12:40
Aula 22

D. J. Benito Gutiérrez, J. L. Aguilar Castro, R. E. Lillo Rodriguez

W. Feng, D. Wozabal, C. Ruiz

S. Rodríguez Santana