Backpropagation through the Void: Optimizing control variates for black-box gradient estimation
Abstract
Gradient-based optimization is the foundation of deep learning and reinforcement learning. Even when the mechanism being optimized is unknown or not differentiable, optimization using high-variance or biased gradient estimates is still often the best strategy. We introduce a general framework for learning low-variance, unbiased gradient estimators for black-box functions of random variables. Our method uses gradients of a neural network trained jointly with model parameters or policies, and is applicable in both discrete and continuous settings. We demonstrate this framework for training discrete latent-variable models. We also give an unbiased, action-conditional extension of the advantage actor-critic reinforcement learning algorithm.
Cite
@article{arxiv.1711.00123,
title = {Backpropagation through the Void: Optimizing control variates for black-box gradient estimation},
author = {Will Grathwohl and Dami Choi and Yuhuai Wu and Geoffrey Roeder and David Duvenaud},
journal= {arXiv preprint arXiv:1711.00123},
year = {2018}
}
Comments
Published at ICLR 2018