TECH·March 20, 2018Variance Reduction for Policy Gradient with Action-Dependent Factorized BaselinesOpenAI