Aug 29, 2023 � We introduce the Relative Gaussian Mechanism (RGM), in which the variance of the noise depends on the norm of the output.
(iv) We show how the Relative Gaussian mechanism can be applied for Private Gradient Descent to provide adaptivity to the gradients' magnitude. We first review�...
Jan 3, 2024 � The Gaussian Mechanism (GM), which consists in adding Gaussian noise to a vector-valued query before releasing it, is a standard privacy�...
The Gaussian mechanism is one differential privacy mechanism commonly used to protect numerical data. However, it may be ill-suited to some applications because�...
Mar 19, 2024 � The Gaussian Mechanism (GM), which con- sists in adding Gaussian noise to a vector- valued query before releasing it, is a stan-.
This repository contains the code for the AISTATS 2024 submission "The Relative Gaussian Mechanism and its Application to Private Gradient Descent".
People also ask
Can we use gradient descent to learn a Gaussian mixture model?
What are the conditions in which gradient descent is applied?
Bellet, The Relative Gaussian Mechanism and its Application to Private Gradient Descent, AISTATS, 2024. [pdf] [arxiv] [hal]. 2023. P. Mangold, Exploiting�...
Leveraging this assumption, we introduce the Relative Gaussian Mechanism (RGM), in which the variance of the noise depends on the norm of the output.
Paul Mangold's 8 research works with 13 citations, including: The Relative Gaussian Mechanism and its Application to Private Gradient Descent.
The Relative Gaussian Mechanism and its Application to Private Gradient Descent � SCAFFLSA: Quantifying and Eliminating Heterogeneity Bias in Federated Linear�...