More Like this

1. Offline reinforcement learning under value and density-ratio realizability: The power of gaps

   Chen, Jinglin ; Jiang, Nan ( August 2022 , Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence)

   We consider a challenging theoretical problem in offline reinforcement learning (RL): obtaining sample-efficiency guarantees with a dataset lacking sufficient coverage, under only realizability-type assumptions for the function approximators. While the existing theory has addressed learning under realizability and under non-exploratory data separately, no work has been able to address both simultaneously (except for a concurrent work which we compare in detail). Under an additional gap assumption, we provide guarantees to a simple pessimistic algorithm based on a version space formed by marginalized importance sampling (MIS), and the guarantee only requires the data to cover the optimal policy and the function classes to realize the optimal value and density-ratio functions. While similar gap assumptions have been used in other areas of RL theory, our work is the first to identify the utility and the novel mechanism of gap assumptions in offline RL with weak function approximation. 

   more » « less
2. Sharper Model-free Reinforcement Learning for Average-reward Markov Decision Processes

   Zhang, Zihan ; Xie, Qiaomin ( July 2023 , Conference on Learning Theory)

   We study model-free reinforcement learning (RL) algorithms for infinite-horizon average-reward Markov decision process (MDP), which is more appropriate for applications that involve continuing operations not divided into episodes. In contrast to episodic/discounted MDPs, theoretical understanding of model-free RL algorithms is relatively inadequate for the average-reward setting. In this paper, we consider both the online setting and the setting with access to a simulator. We develop computationally efficient model-free algorithms that achieve sharper guarantees on regret/sample complexity compared with existing results. In the online setting, we design an algorithm, UCB-AVG, based on an optimistic variant of variance-reduced Q-learning. We show that UCB-AVG achieves a regret bound $\widetilde{O}(S^5A^2sp(h^*)\sqrt{T})$ after $T$ steps, where $S\times A$ is the size of state-action space, and $sp(h^*)$ the span of the optimal bias function. Our result provides the first computationally efficient model-free algorithm that achieves the optimal dependence in $T$ (up to log factors) for weakly communicating MDPs, which is necessary for low regret. In contrast, prior results either are suboptimal in $T$ or require strong assumptions of ergodicity or uniformly mixing of MDPs. In the simulator setting, we adapt the idea of UCB-AVG to develop a model-free algorithm that finds an $\epsilon$-optimal policy with sample complexity $\widetilde{O}(SAsp^2(h^*)\epsilon^{-2} + S^2Asp(h^*)\epsilon^{-1}).$ This sample complexity is near-optimal for weakly communicating MDPs, in view of the minimax lower bound $\Omega(SAsp(^*)\epsilon^{-2})$. Existing work mainly focuses on ergodic MDPs and the results typically depend on $t_{mix},$ the worst-case mixing time induced by a policy. We remark that the diameter $D$ and mixing time $t_{mix}$ are both lower bounded by $sp(h^*)$, and $t_{mix}$ can be arbitrarily large for certain MDPs. On the technical side, our approach integrates two key ideas: learning an $\gamma$-discounted MDP as an approximation, and leveraging reference-advantage decomposition for variance in optimistic Q-learning. As recognized in prior work, a naive approximation by discounted MDPs results in suboptimal guarantees. A distinguishing feature of our method is maintaining estimates of value-difference between state pairs to provide a sharper bound on the variance of reference advantage. We also crucially use a careful choice of the discounted factor $\gamma$ to balance approximation error due to discounting and the statistical learning error, and we are able to maintain a good-quality reference value function with $O(SA)$ space complexity. 

   more » « less
3. Learning Bellman Complete Representations for Offline Policy Evaluation

   Jonathan Chang, Kaiwen Wang ( January 2022 , Proceedings of the 39th International Conference on Machine Learning)

   We study representation learning for Offline Reinforcement Learning (RL), focusing on the important task of Offline Policy Evaluation (OPE). Recent work shows that, in contrast to supervised learning, realizability of the Q-function is not enough for learning it. Two sufficient conditions for sample-efficient OPE are Bellman completeness and coverage. Prior work often assumes that representations satisfying these conditions are given, with results being mostly theoretical in nature. In this work, we propose BCRL, which directly learns from data an approximately linear Bellman complete representation with good coverage. With this learned representation, we perform OPE using Least Square Policy Evaluation (LSPE) with linear functions in our learned representation. We present an end-to-end theoretical analysis, showing that our two-stage algorithm enjoys polynomial sample complexity provided some representation in the rich class considered is linear Bellman complete. Empirically, we extensively evaluate our algorithm on challenging, image-based continuous control tasks from the Deepmind Control Suite. We show our representation enables better OPE compared to previous representation learning methods developed for off-policy RL (e.g., CURL, SPR). BCRL achieve competitive OPE error with the state-of-the-art method Fitted Q-Evaluation (FQE), and beats FQE when evaluating beyond the initial state distribution. Our ablations show that both linear Bellman complete and coverage components of our method are crucial. 

   more » « less
4. Near-Optimal Offline Reinforcement Learning via Double Variance Reduction

   Yin, Ming ; Bai, Yu ; Wang, Yu-Xiang ( December 2021 , Advances in neural information processing systems)

   We consider the problem of offline reinforcement learning (RL) -- a well-motivated setting of RL that aims at policy optimization using only historical data. Despite its wide applicability, theoretical understandings of offline RL, such as its optimal sample complexity, remain largely open even in basic settings such as \emph{tabular} Markov Decision Processes (MDPs). In this paper, we propose Off-Policy Double Variance Reduction (OPDVR), a new variance reduction based algorithm for offline RL. Our main result shows that OPDVR provably identifies an ϵ-optimal policy with O˜(H2/dmϵ2) episodes of offline data in the finite-horizon stationary transition setting, where H is the horizon length and dm is the minimal marginal state-action distribution induced by the behavior policy. This improves over the best known upper bound by a factor of H. Moreover, we establish an information-theoretic lower bound of Ω(H2/dmϵ2) which certifies that OPDVR is optimal up to logarithmic factors. Lastly, we show that OPDVR also achieves rate-optimal sample complexity under alternative settings such as the finite-horizon MDPs with non-stationary transitions and the infinite horizon MDPs with discounted rewards. 

   more » « less
5. COREL: Constrained Reinforcement Learning for Video Streaming ABR Algorithm Design Over mmWave 5G

   https://doi.org/10.1109/CQR59928.2023.10317803  

   Hu, Xinyue ; Ghosh, Arnob ; Liu, Xin ; Zhang, Zhi-Li ; Shroff, Ness ( October 2023 , 2023 IEEE International Workshop Technical Committee on Communications Quality and Reliability (CQR))

   The adaptive bitrate selection (ABR) mechanism, which decides the bitrate for each video chunk is an important part of video streaming. There has been significant interest in developing Reinforcement-Learning (RL) based ABR algorithms because of their ability to learn efficient bitrate actions based on past data and their demonstrated improvements over wired, 3G and 4G networks. However, the Quality of Experience (QoE), especially video stall time, of state-of-the-art ABR algorithms including the RL-based approaches falls short of expectations over commercial mmWave 5G networks, due to widely and wildly fluctuating throughput. These algorithms find optimal policies for a multi-objective unconstrained problem where the policies inherently depend on the predefined weight parameters of the multiple objectives (e.g., bitrate maximization, stall-time minimization). Our empirical evaluation suggests that such a policy cannot adequately adapt to the high variations of 5G throughput, resulting in long stall times. To address these issues, we formulate the ABR selection problem as a constrained Markov Decision Process where the objective is to maximize the QoE subject to a stall-time constraint. The strength of this formulation is that it helps mitigate the stall time while maintaining high bitrates. We propose COREL, a primal-dual actor-critic RL algorithm, which incorporates an additional critic network to estimate stall time compared to existing RL-based approaches and can tune the optimal dual variable or weight to guide the policy towards minimizing stall time. Our experiment results across various commercial mmWave 5G traces reveal that COREL reduces the average stall time by a factor of 4 and the 95th percentile by a factor of 2. 

   more » « less