Offline reinforcement learning -- learning a policy from a batch of data -- is known to be hard for general MDPs. These results motivate the need to look at specific classes of MDPs where offline reinforcement learning might be feasible. In this work, we explore a restricted class of MDPs to obtain guarantees for offline reinforcement learning. The key property, which we call Action Impact Regularity (AIR), is that actions primarily impact a part of the state (an endogenous component) and have limited impact on the remaining part of the state (an exogenous component). AIR is a strong assumption, but it nonetheless holds in a number of real-world domains including financial markets. We discuss algorithms that exploit the AIR property, and provide a theoretical analysis for an algorithm based on Fitted-Q Iteration. Finally, we demonstrate that the algorithm outperforms existing offline reinforcement learning algorithms across different data collection policies in simulated and real world environments where the regularity holds.
Agent-Controller Representations: Principled Offline RL with Rich Exogenous Information
University of Alberta（阿尔伯塔大学）；Amii（阿尔伯塔机器智能研究所）；University of Montreal（蒙特利尔大学）；Quebec AI Institute（魁北克人工智能研究所）；Beijing Institute of Technology（北京理工大学）；Meta；Microsoft Research（微软研究院）；McGill University（麦吉尔大学）
Look Beneath the Surface: Exploiting Fundamental Symmetry for Sample-Efficient Offline RL
Beijing Jiaotong University（北京交通大学）；Tsinghua University（清华大学）；Beijing Key Laboratory of Traffic Data Analysis and Mining（北京交通数据分析与挖掘重点实验室）；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）
Exploiting Generalization in Offline Reinforcement Learning via Unseen State Augmentations
Georgia Institute of Technology（佐治亚理工学院）；Amazon Alexa AI（亚马逊Alexa人工智能）；Allen Institute for Artificial Intelligence（艾伦人工智能研究所）；University of Southern California（南加州大学）
In offline multi-agent reinforcement learning (MARL), agents estimate policies from a given dataset. We study reward-poisoning attacks in this setting where an exogenous attacker modifies the rewards in the dataset before the agents see the dataset. The attacker wants to guide each agent into a nefarious target policy while minimizing the $L^p$ norm of the reward modification. Unlike attacks on single-agent RL, we show that the attacker can install the target policy as a Markov Perfect Dominant Strategy Equilibrium (MPDSE), which rational agents are guaranteed to follow. This attack can be sig…
We study off-policy evaluation and learning from sequential data in a structured class of Markov decision processes that arise from repeated interactions with an exogenous sequence of arrivals with contexts, which generate unknown individual-level responses to agent actions. This model can be thought of as an offline generalization of contextual bandits with resource constraints. We formalize the relevant causal structure of problems such as dynamic personalized pricing and other operations management problems in the presence of potentially high-dimensional user types. The key insight is that …
Arizona State University（亚利桑那州立大学）；The Ohio State University（俄亥俄州立大学）
Existing offline reinforcement learning (RL) methods face a few major challenges, particularly the distributional shift between the learned policy and the behavior policy. Offline Meta-RL is emerging as a promising approach to address these challenges, aiming to learn an informative meta-policy from a collection of tasks. Nevertheless, as shown in our empirical studies, offline Meta-RL could be outperformed by offline single-task RL methods on tasks with good quality of datasets, indicating that a right balance has to be delicately calibrated between "exploring" the out-of-distribution state-a…
Offline reinforcement learning (Offline RL) suffers from the innate distributional shift as it cannot interact with the physical environment during training. To alleviate such limitation, state-based offline RL leverages a learned dynamics model from the logged experience and augments the predicted state transition to extend the data distribution. For exploiting such benefit also on the image-based RL, we firstly propose a generative model, S2P (State2Pixel), which synthesizes the raw pixel of the agent from its corresponding state. It enables bridging the gap between the state and the image d…
Duke University（杜克大学）；Google Research, Brain Team（谷歌研究院大脑团队）
An important problem that arises in reinforcement learning and Monte Carlo methods is estimating quantities defined by the stationary distribution of a Markov chain. In many real-world applications, access to the underlying transition operator is limited to a fixed set of data that has already been collected, without additional interaction with the environment being available. We show that consistent estimation remains possible in this challenging scenario, and that effective estimation can still be achieved in important applications. Our approach is based on estimating a ratio that corrects f…
Reward function is essential in reinforcement learning (RL), serving as the guiding signal to incentivize agents to solve given tasks, however, is also notoriously difficult to design. In many cases, only imperfect rewards are available, which inflicts substantial performance loss for RL agents. In this study, we propose a unified offline policy optimization approach, \textit{RGM (Reward Gap Minimization)}, which can smartly handle diverse types of imperfect rewards. RGM is formulated as a bi-level optimization problem: the upper layer optimizes a reward correction term that performs visitatio…
Layer 6 AI（Layer 6人工智能公司）；University of Toronto（多伦多大学）；Vector Institute（向量研究所）
Skills or low-level policies in reinforcement learning are temporally extended actions that can speed up learning and enable complex behaviours. Recent work in offline reinforcement learning and imitation learning has proposed several techniques for skill discovery from a set of expert trajectories. While these methods are promising, the number K of skills to discover is always a fixed hyperparameter, which requires either prior knowledge about the environment or an additional parameter search to tune it. We first propose a method for offline learning of options (a particular skill framework) …
Inference Time Policy Optimization for Offline RL with Differentiable World Models
Siebel School of Computing and Data Science（计算与数据科学学院）；Department of Computer Sciences（计算机科学系）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；University of Wisconsin Madison（威斯康星大学麦迪逊分校）
This paper considers offline multi-agent reinforcement learning. We propose the strategy-wise concentration principle which directly builds a confidence interval for the joint strategy, in contrast to the point-wise concentration principle that builds a confidence interval for each point in the joint action space. For two-player zero-sum Markov games, by exploiting the convexity of the strategy-wise bonus, we propose a computationally efficient algorithm whose sample complexity enjoys a better dependency on the number of actions than the prior methods based on the point-wise bonus. Furthermore…
We propose SwitchTT, a multi-task extension to Trajectory Transformer but enhanced with two striking features: (i) exploiting a sparsely activated model to reduce computation cost in multi-task offline model learning and (ii) adopting a distributional trajectory value estimator that improves policy performance, especially in sparse reward settings. These two enhancements make SwitchTT suitable for solving multi-task offline reinforcement learning problems, where model capacity is critical for absorbing the vast quantities of knowledge available in the multi-task dataset. More specifically, Swi…
We consider the stochastic multi-armed bandit problem and the contextual bandit problem with historical observations and pre-clustered arms. The historical observations can contain any number of instances for each arm, and the pre-clustering information is a fixed clustering of arms provided as part of the input. We develop a variety of algorithms which incorporate this offline information effectively during the online exploration phase and derive their regret bounds. In particular, we develop the META algorithm which effectively hedges between two other algorithms: one which uses both histori…
