Adaptive Correlated Monte Carlo for Contextual Categorical Sequence Generation
Abstract
Sequence generation models are commonly refined with reinforcement learning over user-defined metrics. However, high gradient variance hinders the practical use of this method. To stabilize this method, we adapt to contextual generation of categorical sequences a policy gradient estimator, which evaluates a set of correlated Monte Carlo (MC) rollouts for variance control. Due to the correlation, the number of unique rollouts is random and adaptive to model uncertainty; those rollouts naturally become baselines for each other, and hence are combined to effectively reduce gradient variance. We also demonstrate the use of correlated MC rollouts for binary-tree softmax models, which reduce the high generation cost in large vocabulary scenarios by decomposing each categorical action into a sequence of binary actions. We evaluate our methods on both neural program synthesis and image captioning. The proposed methods yield lower gradient variance and consistent improvement over related baselines.
Cite
@article{arxiv.1912.13151,
title = {Adaptive Correlated Monte Carlo for Contextual Categorical Sequence Generation},
author = {Xinjie Fan and Yizhe Zhang and Zhendong Wang and Mingyuan Zhou},
journal= {arXiv preprint arXiv:1912.13151},
year = {2020}
}
Comments
ICLR 2020 (updated to fix a typo in Algorithm 1)