Adaptive Correlated Monte Carlo for Contextual Categorical Sequence Generation
arXiv:1912.13151
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.
arXiv commentICLR 2020 (updated to fix a typo in Algorithm 1)
References in corpus (21)
- Efficient Estimation of Word Representations in Vector Space
- Auto-Encoding Variational Bayes
- Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
- Categorical Reparameterization with Gumbel-Softmax
- Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks
- A Deep Reinforced Model for Abstractive Summarization
- The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
- Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning
- Improved Image Captioning via Policy Gradient optimization of SPIDEr
- Professor Forcing: A New Algorithm for Training Recurrent Networks
- An Actor-Critic Algorithm for Sequence Prediction
- Neural Variational Inference and Learning in Belief Networks
- REBAR: Low-variance, unbiased gradient estimates for discrete latent variable models
- Variational Bayesian Inference with Stochastic Search
- Reinforcement Learning Neural Turing Machines - Revised
- RobustFill: Neural Program Learning under Noisy I/O
- Actor-Critic Sequence Training for Image Captioning
- Classes for Fast Maximum Entropy Training
- ARM: Augment-REINFORCE-Merge Gradient for Stochastic Binary Networks
- ARSM: Augment-REINFORCE-Swap-Merge Estimator for Gradient Backpropagation Through Categorical Variables
- Learning to Screen for Fast Softmax Inference on Large Vocabulary Neural Networks