More Like this

1. 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
2. Optimistic Posterior Sampling for Reinforcement Learning: Worst-Case Regret Bounds

   https://doi.org/10.1287/moor.2022.1266  

   Agrawal, Shipra; Jia, Randy (May 2022, Mathematics of Operations Research)

   We present an algorithm based on posterior sampling (aka Thompson sampling) that achieves near-optimal worst-case regret bounds when the underlying Markov decision process (MDP) is communicating with a finite, although unknown, diameter. Our main result is a high probability regret upper bound of [Formula: see text] for any communicating MDP with S states, A actions, and diameter D. Here, regret compares the total reward achieved by the algorithm to the total expected reward of an optimal infinite-horizon undiscounted average reward policy in time horizon T. This result closely matches the known lower bound of [Formula: see text]. Our techniques involve proving some novel results about the anti-concentration of Dirichlet distribution, which may be of independent interest. 

   more » « less
3. Provable Policy Gradient Methods for Average-Reward Markov Potential Games

   Cheng, Min; Zhou, Ruida; Kumar, P R; Tian, Chao (May 2024, Proceedings of the 27th International Conference on Artificial Intelligence and Statistics (AISTATS) 2024)

   We study Markov potential games under the infinite horizon average reward criterion. Most previous studies have been for discounted rewards. We prove that both algorithms based on independent policy gradient and independent natural policy gradient converge globally to a Nash equilibrium for the average reward criterion. To set the stage for gradient-based methods, we first establish that the average reward is a smooth function of policies and provide sensitivity bounds for the differential value functions, under certain conditions on ergodicity and the second largest eigenvalue of the underlying Markov decision process (MDP). We prove that three algorithms, policy gradient, proximal-Q, and natural policy gradient (NPG), converge to an ϵ-Nash equilibrium with time complexity O(1ϵ2), given a gradient/differential Q function oracle. When policy gradients have to be estimated, we propose an algorithm with ~O(1mins,aπ(a|s)δ) sample complexity to achieve δ approximation error w.r.t~the ℓ2 norm. Equipped with the estimator, we derive the first sample complexity analysis for a policy gradient ascent algorithm, featuring a sample complexity of ~O(1/ϵ5). Simulation studies are presented. 

   more » « less
4. Near-Optimal Reinforcement Learning with Self-Play under Adaptivity Constraints

   Qiao, Dan; Wang, Yu-Xiang (July 2024, Proceedings of Machine Learning Research)

   We study the problem of multi-agent reinforcement learning (MARL) with adaptivity constraints -- a new problem motivated by real-world applications where deployments of new policies are costly and the number of policy updates must be minimized. For two-player zero-sum Markov Games, we design a (policy) elimination based algorithm that achieves a regret of O˜(H3S2ABK‾‾‾‾‾‾‾‾‾‾√), while the batch complexity is only O(H+loglogK). In the above, S denotes the number of states, A,B are the number of actions for the two players respectively, H is the horizon and K is the number of episodes. Furthermore, we prove a batch complexity lower bound Ω(HlogAK+loglogK) for all algorithms with O˜(K‾‾√) regret bound, which matches our upper bound up to logarithmic factors. As a byproduct, our techniques naturally extend to learning bandit games and reward-free MARL within near optimal batch complexity. To the best of our knowledge, these are the first line of results towards understanding MARL with low adaptivity. 

   more » « less
5. Towards Tighter Space Bounds for Counting Triangles and Other Substructures in Graph Streams

   Bera, Suman K; Chakrabarti, Amit (March 2017, 34th Symposium on Theoretical Aspects of Computer Science)

   We revisit the much-studied problem of space-efficiently estimating the number of triangles in a graph stream, and extensions of this problem to counting fixed-sized cliques and cycles, obtaining a number of new upper and lower bounds. For the important special case of counting triangles, we give a $$4$$-pass, $$(1\pm\varepsilon)$$-approximate, randomized algorithm that needs at most $$\widetilde{O}(\varepsilon^{-2}\cdot m^{3/2}/T)$$ space, where $$m$$ is the number of edges and $$T$$ is a promised lower bound on the number of triangles. This matches the space bound of a very recent algorithm (McGregor et al., PODS 2016), with an arguably simpler and more general technique. We give an improved multi-pass lower bound of $$\Omega(\min\{m^{3/2}/T, m/\sqrt{T}\})$$, applicable at essentially all densities $$\Omega(n) \le m \le O(n^2)$$. We also prove other multi-pass lower bounds in terms of various structural parameters of the input graph. Together, our results resolve a couple of open questions raised in recent work (Braverman et al., ICALP 2013). Our presentation emphasizes more general frameworks, for both upper and lower bounds. We give a sampling algorithm for counting arbitrary subgraphs and then improve it via combinatorial means in the special cases of counting odd cliques and odd cycles. Our results show that these problems are considerably easier in the cash-register streaming model than in the turnstile model, where previous work had focused (Manjunath et al., ESA 2011; Kane et al., ICALP 2012). We use Tur{\'a}n graphs and related gadgets to derive lower bounds for counting cliques and cycles, with triangle-counting lower bounds following as a corollary. 

   more » « less