Contraction of $E_γ$-Divergence and Its Applications to Privacy

We investigate the contraction coefficients derived from strong data processing inequalities for the $E_γ$-divergence. By generalizing the celebrated Dobrushin's coefficient from total variation distance to $E_γ$-divergence, we derive a closed-form expression for the contraction of $E_γ$-divergence....

Descripción completa

Detalles Bibliográficos

Autores principales:	Asoodeh, Shahab, Diaz, Mario, Calmon, Flavio P.
Formato:	Artículo
Publicado:	2020
Materias:	Information Theory Cryptography and Security Machine Learning
Acceso en línea:	https://arxiv.org/abs/2012.11035 http://arxiv.org/abs/2012.11035
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!