Model-free Posterior Sampling via Learning Rate Randomization

In this paper, we introduce Randomized Q-learning (RandQL), a novel randomized model-free algorithm for regret minimization in episodic Markov Decision Processes (MDPs). To the best of our knowledge, RandQL is the first tractable model-free posterior sampling-based algorithm. We analyze the performance of RandQL in both tabular and non-tabular metric space settings. In tabular MDPs, RandQL achieves a regret bound of order $\widetilde{\mathcal{O}}(\sqrt{H^{5}SAT})$, where $H$ is the planning horizon, $S$ is the number of states, $A$ is the number of actions, and $T$ is the number of episodes. For a metric state-action space, RandQL enjoys a regret bound of order $\widetilde{\mathcal{O}}(H^{5/2} T^{(d_z+1)/(d_z+2)})$, where $d_z$ denotes the zooming dimension. Notably, RandQL achieves optimistic exploration without using bonuses, relying instead on a novel idea of learning rate randomization. Our empirical study shows that RandQL outperforms existing approaches on baseline exploration environments.

Demonstration-Regularized RL

Incorporating expert demonstrations has empirically helped to improve the sample efficiency of reinforcement learning (RL). This paper quantifies theoretically to what extent this extra information reduces RL's sample complexity. In particular, we study the demonstration-regularized reinforcement learning that leverages the expert demonstrations by KL-regularization for a policy learned by behavior cloning. Our findings reveal that using $N^{\mathrm{E}}$ expert demonstrations enables the identification of an optimal policy at a sample complexity of order $\widetilde{\mathcal{O}}(\mathrm{Poly}(S,A,H)/(\varepsilon^2 N^{\mathrm{E}}))$ in finite and $\widetilde{\mathcal{O}}(\mathrm{Poly}(d,H)/(\varepsilon^2 N^{\mathrm{E}}))$ in linear Markov decision processes, where $\varepsilon$ is the target precision, $H$ the horizon, $A$ the number of action, $S$ the number of states in the finite case and $d$ the dimension of the feature space in the linear case. As a by-product, we provide tight convergence guarantees for the behaviour cloning procedure under general assumptions on the policy classes. Additionally, we establish that demonstration-regularized methods are provably efficient for reinforcement learning from human feedback (RLHF). In this respect, we provide theoretical evidence showing the benefits of KL-regularization for RLHF in tabular and linear MDPs. Interestingly, we avoid pessimism injection by employing computationally feasible regularization to handle reward estimation uncertainty, thus setting our approach apart from the prior works.

Fast Rates for Maximum Entropy Exploration

We consider the reinforcement learning (RL) setting, in which the agent has to act in unknown environment driven by a Markov Decision Process (MDP) with sparse or even reward free signals. In this situation, exploration becomes the main challenge. In this work, we study the maximum entropy exploration problem of two different types. The first type is visitation entropy maximization that was previously considered by Hazan et al. (2019) in the discounted setting. For this type of exploration, we propose an algorithm based on a game theoretic representation that has $\widetilde{\mathcal{O}}(H^3 S^2 A / \varepsilon^2)$ sample complexity thus improving the $\varepsilon$-dependence of Hazan et al. (2019), where $S$ is a number of states, $A$ is a number of actions, $H$ is an episode length, and $\varepsilon$ is a desired accuracy. The second type of entropy we study is the trajectory entropy. This objective function is closely related to the entropy-regularized MDPs, and we propose a simple modification of the UCBVI algorithm that has a sample complexity of order $\widetilde{\mathcal{O}}(1/\varepsilon)$ ignoring dependence in $S, A, H$. Interestingly enough, it is the first theoretical result in RL literature establishing that the exploration problem for the regularized MDPs can be statistically strictly easier (in terms of sample complexity) than for the ordinary MDPs.

When Combinatorial Thompson Sampling meets Approximation Regret

We study the Combinatorial Thompson Sampling policy (CTS) for combinatorial multi-armed bandit problems (CMAB), within an approximation regret setting. Although CTS has attracted a lot of interest, it has a drawback that other usual CMAB policies do not have when considering non-exact oracles: for some oracles, CTS has a poor approximation regret (scaling linearly with the time horizon $T$) [Wang and Chen, 2018]. A study is then necessary to discriminate the oracles on which CTS could learn. This study was started by Kong et al. [2021]: they gave the first approximation regret analysis of CTS for the greedy oracle, obtaining an upper bound of order $\mathcal{O}(\log(T)/\Delta^2)$, where $\Delta$ is some minimal reward gap. In this paper, our objective is to push this study further than the simple case of the greedy oracle. We provide the first $\mathcal{O}(\log(T)/\Delta)$ approximation regret upper bound for CTS, obtained under a specific condition on the approximation oracle, allowing a reduction to the exact oracle analysis. We thus term this condition REDUCE2EXACT, and observe that it is satisfied in many concrete examples. Moreover, it can be extended to the probabilistically triggered arms setting, thus capturing even more problems, such as online influence maximization.

Statistical Efficiency of Thompson Sampling for Combinatorial Semi-Bandits

We investigate stochastic combinatorial multi-armed bandit with semi-bandit feedback (CMAB). In CMAB, the question of the existence of an efficient policy with an optimal asymptotic regret (up to a factor poly-logarithmic with the action size) is still open for many families of distributions, including mutually independent outcomes, and more generally the multivariate sub-Gaussian family. We propose to answer the above question for these two families by analyzing variants of the Combinatorial Thompson Sampling policy (CTS). For mutually independent outcomes in $[0,1]$, we propose a tight analysis of CTS using Beta priors. We then look at the more general setting of multivariate sub-Gaussian outcomes and propose a tight analysis of CTS using Gaussian priors. This last result gives us an alternative to the Efficient Sampling for Combinatorial Bandit policy (ESCB), which, although optimal, is not computationally efficient.

Active Linear Regression

We consider the problem of active linear regression where a decision maker has to choose between several covariates to sample in order to obtain the best estimate $\hat{\beta}$ of the parameter $\beta^{\star}$ of the linear model, in the sense of minimizing $\mathbb{E} \lVert\hat{\beta}-\beta^{\star}\rVert^2$. Using bandit and convex optimization techniques we propose an algorithm to define the sampling strategy of the decision maker and we compare it with other algorithms. We provide theoretical guarantees of our algorithm in different settings, including a $\mathcal{O}(T^{-2})$ regret bound in the case where the covariates form a basis of the feature space, generalizing and improving existing results. Numerical experiments validate our theoretical findings.

Exploiting Structure of Uncertainty for Efficient Combinatorial Semi-Bandits

We improve the efficiency of algorithms for stochastic \emph{combinatorial semi-bandits}. In most interesting problems, state-of-the-art algorithms take advantage of structural properties of rewards, such as \emph{independence}. However, while being minimax optimal in terms of regret, these algorithms are intractable. In our paper, we first reduce their implementation to a specific \emph{submodular maximization}. Then, in case of \emph{matroid} constraints, we design adapted approximation routines, thereby providing the first efficient algorithms that exploit the reward structure. In particular, we improve the state-of-the-art efficient gap-free regret bound by a factor $\sqrt{k}$, where $k$ is the maximum action size. Finally, we show how our improvement translates to more general \emph{budgeted combinatorial semi-bandits}.

Finding the Bandit in a Graph: Sequential Search-and-Stop

We consider the problem where an agent wants to find a hidden object that is randomly located in some vertex of a directed acyclic graph (DAG) according to a fixed but possibly unknown distribution. The agent can only examine vertices whose in-neighbors have already been examined. In scheduling theory, this problem is denoted by $1|prec|\sum w_jC_j$ [Graham1979]. However, in this paper we address a learning setting where we allow the agent to stop before having found the object and restart searching on a new independent instance of the same problem. The goal is to maximize the total number of hidden objects found under a time constraint. The agent can thus skip an instance after realizing that it would spend too much time on it. Our contributions are both to the search theory and multi-armed bandits. If the distribution is known, we provide a quasi-optimal greedy strategy with the help of known computationally efficient algorithms for solving $1|prec|\sum w_jC_j$ under some assumption on the DAG. If the distribution is unknown, we show how to sequentially learn it and, at the same time, act near-optimally in order to collect as many hidden objects as possible. We provide an algorithm, prove theoretical guarantees, and empirically show that it outperforms the naive baseline.