DifferentiableExpectations.jl

This package revolves around functions defined as expectations:

$$F(\theta) = \mathbb{E}_{p_\theta}[f(X)]$$

It allows the computation of approximate derivatives with respect to $\theta$ thanks to Monte-Carlo samples.

For more details, refer to the following paper:

Monte-Carlo Gradient Estimation in Machine Learning, Mohamed et al. (2020)