Gradient Estimation with Discrete Stein Operators
This addresses a key bottleneck in machine learning for problems involving discrete distributions, such as generative modeling, by improving gradient estimation efficiency.
The paper tackled the problem of high variance in gradient estimation for discrete distributions by introducing a variance reduction technique based on Stein operators, which led to substantially lower variance in benchmark tasks like training binary variational autoencoders compared to state-of-the-art estimators.
Gradient estimation -- approximating the gradient of an expectation with respect to the parameters of a distribution -- is central to the solution of many machine learning problems. However, when the distribution is discrete, most common gradient estimators suffer from excessive variance. To improve the quality of gradient estimation, we introduce a variance reduction technique based on Stein operators for discrete distributions. We then use this technique to build flexible control variates for the REINFORCE leave-one-out estimator. Our control variates can be adapted online to minimize variance and do not require extra evaluations of the target function. In benchmark generative modeling tasks such as training binary variational autoencoders, our gradient estimator achieves substantially lower variance than state-of-the-art estimators with the same number of function evaluations.