• Corpus ID: 233423379

Autoregressive Dynamics Models for Offline Policy Evaluation and Optimization

@article{Zhang2021AutoregressiveDM,
  title={Autoregressive Dynamics Models for Offline Policy Evaluation and Optimization},
  author={Michael R. Zhang and Tom Le Paine and Ofir Nachum and Cosmin Paduraru and G. Tucker and Ziyun Wang and Mohammad Norouzi},
  journal={ArXiv},
  year={2021},
  volume={abs/2104.13877}
}

• Michael R. Zhang, T. Paine, +4 authors Mohammad Norouzi
• Published 28 April 2021
• Computer Science
• ArXiv

Standard dynamics models for continuous control make use of feedforward computation to predict the conditional distribution of next state and reward given current state and action using a multivariate Gaussian with a diagonal covariance structure. This modeling choice assumes that different dimensions of the next state and reward are conditionally independent given the current state and action and may be driven by the fact that fully observable physics-based simulation environments entail deterministic transition dynamics. In this paper, we challenge this conditional independence assumption and propose a family of expressive autoregressive dynamics models that generate different dimensions of the next state and reward sequentially conditioned on previous dimensions. We demonstrate that autoregressive dynamics models indeed outperform standard feedforward models in log-likelihood on heldout transitions. Furthermore, we compare different model-based and model-free off-policy evaluation (OPE) methods on RL Unplugged, a suite of offline MuJoCo datasets, and find that autoregressive dynamics models consistently outperform all baselines, achieving a new state-of-the-art. Finally, we show that autoregressive dynamics models are useful for offline policy optimization by serving as a way to enrich the replay buffer through data augmentation and improving performance using model-based planning.