How moving randomness outside the computation graph turns noisy gradient estimators into low-variance, differentiable ones
The post Reparameterization Tricks: Variance Reduction by Smarter Gradients appeared first on Towards Data Science.
Science News
