While standard bandit algorithms sometimes incur high regret, their performance can be greatly improved by "warm starting" with historical data. Unfortunately, how best to incorporate historical data is unclear: naively initializing reward estimates using all historical samples can suffer from spurious data and imbalanced data coverage, leading to computational and storage issues - particularly in continuous action spaces. We address these two challenges by proposing Artificial Replay, a meta-algorithm for incorporating historical data into any arbitrary base bandit algorithm. Artificial Replay uses only a subset of the historical data as needed to reduce computation and storage. We show that for a broad class of base algorithms that satisfy independence of irrelevant data (IIData), a novel property that we introduce, our method achieves equal regret as a full warm-start approach while potentially using only a fraction of the historical data. We complement these theoretical results with a case study of $K$-armed and continuous combinatorial bandit algorithms, including on a green security domain using real poaching data, to show the practical benefits of Artificial Replay in achieving optimal regret alongside low computational and storage costs.
We develop a reinforcement learning (RL) framework for applications that deal with sequential decisions and exogenous uncertainty, such as resource allocation and inventory management. In these applications, the uncertainty is only due to exogenous variables like future demands. A popular approach is to predict the exogenous variables using historical data and then plan with the predictions. However, this indirect approach requires high-fidelity modeling of the exogenous process to guarantee good downstream decision-making, which can be impractical when the exogenous process is complex. In this work we propose an alternative approach based on hindsight learning which sidesteps modeling the exogenous process. Our key insight is that, unlike Sim2Real RL, we can revisit past decisions in the historical data and derive counterfactual consequences for other actions in these applications. Our framework uses hindsight-optimal actions as the policy training signal and has strong theoretical guarantees on decision-making performance. We develop an algorithm using our framework to allocate compute resources for real-world Microsoft Azure workloads. The results show our approach learns better policies than domain-specific heuristics and Sim2Real RL baselines.
Discretization based approaches to solving online reinforcement learning problems have been studied extensively in practice on applications ranging from resource allocation to cache management. Two major questions in designing discretization-based algorithms are how to create the discretization and when to refine it. While there have been several experimental results investigating heuristic solutions to these questions, there has been little theoretical treatment. In this paper we provide a unified theoretical analysis of tree-based hierarchical partitioning methods for online reinforcement learning, providing model-free and model-based algorithms. We show how our algorithms are able to take advantage of inherent structure of the problem by providing guarantees that scale with respect to the 'zooming dimension' instead of the ambient dimension, an instance-dependent quantity measuring the benignness of the optimal $Q_h^\star$ function. Many applications in computing systems and operations research requires algorithms that compete on three facets: low sample complexity, mild storage requirements, and low computational burden. Our algorithms are easily adapted to operating constraints, and our theory provides explicit bounds across each of the three facets. This motivates its use in practical applications as our approach automatically adapts to underlying problem structure even when very little is known a priori about the system.
We introduce the technique of adaptive discretization to design efficient model-based episodic reinforcement learning algorithms in large (potentially continuous) state-action spaces. Our algorithm is based on optimistic one-step value iteration extended to maintain an adaptive discretization of the space. From a theoretical perspective, we provide worst-case regret bounds for our algorithm, which are competitive compared to the state-of-the-art model-based algorithms; moreover, our bounds are obtained via a modular proof technique, which can potentially extend to incorporate additional structure on the problem. From an implementation standpoint, our algorithm has much lower storage and computational requirements, due to maintaining a more efficient partition of the state and action spaces. We illustrate this via experiments on several canonical control problems, which shows that our algorithm empirically performs significantly better than fixed discretization in terms of both faster convergence and lower memory usage. Interestingly, we observe empirically that while fixed-discretization model-based algorithms vastly outperform their model-free counterparts, the two achieve comparable performance with adaptive discretization.
We present an efficient algorithm for model-free episodic reinforcement learning on large (potentially continuous) state-action spaces. Our algorithm is based on a novel $Q$-learning policy with adaptive data-driven discretization. The central idea is to maintain a finer partition of the state-action space in regions which are frequently visited in historical trajectories, and have higher payoff estimates. We demonstrate how our adaptive partitions take advantage of the shape of the optimal $Q$-function and the joint space, without sacrificing the worst-case performance. In particular, we recover the regret guarantees of prior algorithms for continuous state-action spaces, which additionally require either an optimal discretization as input, and/or access to a simulation oracle. Moreover, experiments demonstrate how our algorithm automatically adapts to the underlying structure of the problem, resulting in much better performance compared both to heuristics and $Q$-learning with uniform discretization.