Core Library#
Higher-level APIs#
End-to-end training loop for differentially private training. |
|
Module for defining DP Execution Plans. |
|
API for adding DP-SGD to a Keras model. |
Individual Components#
Functions for creating DpEvent objects for common JAX Privacy mechanisms. |
|
API and implementations for batch selection strategies. |
|
Utilities for clipping function outputs and aggregating across a batch. |
|
Implementations of optax.GradientTransformations that add noise to gradients. |
|
Augmented gradient transformations for differentially private training. |
|
Library for empirical privacy auditing/estimation. |
|
Public API for matrix factorization. |
Experimental Modules#
Experimental utilities for handling variable batch sizes. |
|
Monte Carlo accounting module. |