Two new algorithms improve model-free RL for infinite-horizon MDPs.
problem Learning in infinite-horizon average-reward MDPs.
method Two model-free algorithms for infinite-horizon average-reward MDPs.
result Improved regret bounds for model-free RL in MDPs.
Improved Q-learning with UCB for infinite-horizon MDPs with sample complexity bound.
problem Sample efficiency of Q-learning with UCB for infinite-horizon MDPs.
method Adapted Q-learning with UCB-exploration bonus for infinite-horizon MDPs without generative model.
result Sample complexity of exploration is bounded by O({ rac{SA}{ε^2(1-γ)^7}}), improving previous results.
A new model-free algorithm achieves near-optimal regret for infinite-horizon MDPs.
problem Model-free reinforcement learning for infinite-horizon average-reward MDPs.
method Exploration Enhanced Q-learning (EE-QL) for weakly communicating MDPs.
result Achieves O ( T ) O(\sqrt{T}) O ( T ) regret bound for general weakly communicating MDPs. UCRL2-VTR achieves nearly optimal regret for learning MDPs with linear function approximation.
problem Learning infinite-horizon average-reward MDPs with linear function approximation.
method UCRL2-VTR algorithm with Bernstein-type bonus.
result Achieves a regret of i l d e O ( d D T ) ilde{O}(d\sqrt{DT}) i l d e O ( d D T ) with matching lower bound. Paper proposes an efficient online learning method using an offline dataset for infinite horizon MDPs.
problem Efficient online reinforcement learning in infinite horizon MDPs with an unknown expert policy.
method Bayesian approach to model the expert's policy and minimize cumulative regret.
result Upper bound on regret of i l d e O ( T ) ilde{O}(\sqrt{T}) i l d e O ( T ) for the Informed PSRL algorithm. New algorithm reduces reinforcement learning regret to sqrt(T) without strong dynamics assumptions.
problem Infinite-horizon average-reward reinforcement learning with linear MDPs.
method Approximate by discounted-reward MDPs and apply optimistic value iteration.
result Achieves O(sqrt(T)) regret with polynomial complexity.
New algorithms for learning MDPs with linear approximations in infinite-horizon settings.
problem Learning infinite-horizon average-reward MDPs with linear function approximation.
method Optimism principle, adversarial linear bandits, Natural Policy Gradient.
result Efficient algorithms with optimal or near-optimal regret bounds.
New algorithms reduce dynamic regret in online MDPs with changing losses.
problem Online MDPs with adversarial loss changes and known transitions.
method Dynamic regret measure, novel ensemble algorithms for three models.
result Provably optimal dynamic regret bounds for episodic SSP, improved bounds for predictable environments.
Actor-Critic method achieves optimal regret for unichain MDPs.
problem Scalable regret analysis for infinite-horizon average-reward MDPs.
method NAC-B, a Natural Actor-Critic with batching.
result Order-optimal regret of i l d e O ( T ) ilde{O}(\sqrt{T}) i l d e O ( T ) in infinite-horizon average-reward MDPs. Improved RL algorithm with linear MDPs for offline learning with partial data coverage.
problem Efficient offline RL with linear MDPs under partial data coverage.
method Primal-dual algorithm with O ( ε − 2 ) O(ε^{-2}) O ( ε − 2 ) sample complexity. result First computationally efficient algorithm with O ( ε − 2 ) O(ε^{-2}) O ( ε − 2 ) sample complexity for offline RL with linear MDPs under partial data coverage. New algorithms learn MDPs with better regret bounds using generative sampling.
problem Learning MDPs with optimal policies under uncertainty.
method Hybrid exploration-generative RL model, classical and quantum algorithms.
result Quantum algorithms achieve poly log T \operatorname{poly}\log{T} poly log T regret for infinite-horizon MDPs. New algorithm reduces regret in infinite MDPs with optimal variance-dependent bounds.
problem Infinite horizon MDPs lack optimal algorithms with low regret.
method Developed a UCB-style algorithm for average-reward and γ-regret.
result Achieved optimal variance-dependent regret bounds for both objectives.
This paper is concerned with offline reinforcement learning (RL), which learns using pre-collected data without further exploration. Effective offline RL would be able to accommodate distribution shift and limited data coverage. However, prior algorithms or analyses either suffer from suboptimal sample complexities or …
This paper improves Thompson Sampling for complex decision-making problems.
problem Learning in infinite-horizon discounted decision processes with unknown parameters.
method Developed a general canonical probability space and new metrics for analyzing adaptive learning algorithms.
result Thompson Sampling achieves complete learning in complex decision-making problems.
A new parallel algorithm for learning optimal policies in MDPs with low communication costs.
problem Learning optimal policies for infinite-horizon MDPs.
method Primal-Dual Stochastic Mirror Descent for convex programming problems with inexact constraints.
result First parallel algorithm for average-reward MDPs with generative model and low communication costs.
New method for optimistic planning in MDPs using regularization.
problem Optimistic planning in infinite-horizon discounted MDPs.
method Regularized dynamic programming for approximate value iteration.
result Achieves near-optimal statistical guarantees in learning policies.
NVMDP framework tackles non-stationary MDPs with varying discount rates.
problem Challenges in non-stationary environments and infinite-horizon formulations for reinforcement learning.
method Introduces NVMDP framework that accommodates non-stationarity and varying discount rates.
result NVMDPs provide a flexible mechanism to shape optimal policies without altering state or action spaces.
We study the online estimation of the optimal policy of a Markov decision process (MDP). We propose a class of Stochastic Primal-Dual (SPD) methods which exploit the inherent minimax duality of Bellman equations. The SPD methods update a few coordinates of the value and policy estimates as a new state transition is obs…
Logarithmic regret for continuous-time reinforcement learning.
problem Continuous-time Markov decision processes with unknown transition probabilities and holding times.
method Upper confidence reinforcement learning, mean holding time estimation, stochastic comparison of point processes.
result Logarithmic regret bound achieved in finite time.
Unified framework for solving MDPs with stochastic mirror descent.
problem Approximately solving infinite-horizon Markov decision processes (MDPs).
method Primal-dual stochastic mirror descent for MDPs with a unified framework.
result Computes ε-optimal policies with expected samples for both average-reward and discounted MDPs.
Paper overcomes sample size barrier in reinforcement learning with generative models.
problem Sample efficiency in reinforcement learning with generative models.
method Developed two algorithms to certify minimax optimality of sample complexity.
result Achieved minimax-optimal guarantees for a wide range of sample sizes.
Paper develops IV method for consistent OPE in confounded MDPs.
problem Confounding variables in observational data affect OPE effectiveness.
method Instrumental variable approach for consistent off-policy evaluation.
result IV method enables correct identification of target policy's value.
TensorPlan algorithm finds δ-optimal policies with poly ( H , d ) (H,d) ( H , d ) queries under linearly realizable state-value function.
problem Efficient planning in MDPs with linearly realizable state-value function.
method TensorPlan algorithm using poly ( ( d H / δ ) A ) ((dH/δ)^A) (( d H / δ ) A ) simulator queries. result First algorithm with polynomial query complexity using only linear-realizability of a single competing value function.
This is a brief technical note to clarify some of the issues with applying the application of the algorithm posterior sampling for reinforcement learning (PSRL) in environments without fixed episodes. In particular, this paper aims to: - Review some of results which have been proven for finite horizon MDPs (Osband et a…
Paper analyzes convergence of dynamic policy gradient for MDPs, improving performance in finite-time problems.
problem Optimal policies in finite-time MDPs are not stationary and require epoch-specific training.
method Introduces dynamic policy gradient combining dynamic programming and policy gradient, analyzes convergence for softmax parametrisation.
result Dynamic policy gradient training exploits finite-time structure, leading to better convergence bounds.
RL approach for target tracking with unknown dynamics and sensor control.
problem Tracking an unknown target with sensor control.
method Track-MDP formulation for RL, compared with POMDP.
result Optimal RL policy tracks all target paths with certainty.
Study shows policy gradient convergence for entropy-regularized MDPs with neural nets in mean-field regime.
problem Global convergence of policy gradient for entropy-regularized MDPs with neural network approximation.
method Softmax policy with neural network approximation in mean-field regime, gradient flow in 2-Wasserstein metric, exponential convergence under sufficient regularization.
result Gradient flow converges exponentially fast to the unique stationary solution under sufficient regularization.
New RL theory reduces sample complexity for mixing MDPs.
problem Optimal sample complexity for reinforcement learning in mixing MDPs.
method Regeneration-type ideas to analyze mixing times.
result Optimal sample complexity depends on mixing time, not just discount factor.
Algorithm improves reinforcement learning in MDPs with partial order policies.
problem Improving reinforcement learning in MDPs with partial order policies.
method Epoch-based reinforcement learning algorithm leveraging a partial order over policy class.
result Achieves an O ( w log ( ∣ Θ ∣ ) T ) O(\sqrt{w \log(|Θ|) T}) O ( w log ( ∣Θ∣ ) T ) regret bound, independent of state and action space sizes. We characterize value functions in partially observable MDPs as semi-algebraic sets.
problem Understanding feasible value functions in partially observable Markov decision processes.
method Characterization of feasible value functions as semi-algebraic sets defined by polynomial inequalities.
result The feasible set of value functions in POMDPs is a semi-algebraic set, not a polytope as in MDPs.
New approach turns optimal stationary RL into non-stationary RL without prior knowledge.
problem Optimal RL in non-stationary environments without prior knowledge of non-stationarity.
method Black-box reduction of optimal stationary RL algorithms to non-stationary RL.
result Achieves optimal dynamic regret bounds in various RL settings.
We consider model-free reinforcement learning for infinite-horizon discounted Markov Decision Processes (MDPs) with a continuous state space and unknown transition kernel, when only a single sample path under an arbitrary policy of the system is available. We consider the Nearest Neighbor Q-Learning (NNQL) algorithm to…
Improved sample complexity for actor-critic algorithms in MDPs.
problem Achieving optimal policies with limited data in reinforcement learning.
method Single-timescale actor-critic with STORM (STOchastic Recursive Momentum) and a sample buffer.
result Optimal sample complexity of O ( ε − 2 ) O(ε^{-2}) O ( ε − 2 ) for ε ε ε -optimal policies. The paper studies reward concentration in MDPs, covering asymptotic and non-asymptotic settings.
problem Reward concentration in Markov Decision Processes (MDPs).
method Unified approach to reward concentration in MDPs, including asymptotic and non-asymptotic bounds.
result Rate-equivalent definitions of regret for learning policies.
New algorithm learns optimal policies with minimal memory and time.
problem Learning optimal policies in discounted MDPs with short burn-in time.
method Variance reduction and adaptive policy switching.
result First regret-optimal model-free algorithm with low burn-in time.
Improved online Q-learning for MDPs with concentration bounds.
problem Online Q-learning in infinite-horizon discounted MDPs with sublinear regret for large gaps.
method Smoothed ε n ε_n ε n -Greedy exploration scheme combining ε n ε_n ε n -greedy and Boltzmann exploration, analyzed using concentration bounds for contractive Markovian stochastic approximation. result Near- i l d e O ( N 9 / 10 ) ilde{O}(N^{9/10}) i l d e O ( N 9/10 ) regret bound for Smoothed ε n ε_n ε n -Greedy exploration scheme. New RL method handles hidden actions in offline learning.
problem Learning from unseen actions in real-world RL datasets.
method LURE (Learning from the Unseen: Robust Estimator) method using next-state variable as proxy.
result Valid statistical inference and improved RL conclusions with hidden actions.
Recently, there has been significant progress in understanding reinforcement learning in discounted infinite-horizon Markov decision processes (MDPs) by deriving tight sample complexity bounds. However, in many real-world applications, an interactive learning agent operates for a fixed or bounded period of time, for ex…
To overcome the curse of dimensionality and curse of modeling in Dynamic Programming (DP) methods for solving classical Markov Decision Process (MDP) problems, Reinforcement Learning (RL) algorithms are popular. In this paper, we consider an infinite-horizon average reward MDP problem and prove the optimality of the th…
Paper develops a method to estimate value of a policy in confounded MDPs.
problem Estimating value of a policy in the presence of unmeasured confounders.
method Uses auxiliary variables to identify target policy's value in a confounded MDP.
result Develops an off-policy value estimator robust to model misspecification.
Efficiently selects seed nodes to maximize content influence in unknown social networks.
problem Maximizing content spread in social networks with unknown network model.
method Formulated as an infinite-horizon discounted MDP, uses model-based reinforcement learning to select seed users adaptively.
result Established a regret bound of O ~ ( T ) \widetilde O(\sqrt{T}) O ( T ) for the algorithm. This paper considers a transmission control problem in network-coded two-way relay channels (NC-TWRC), where the relay buffers random symbol arrivals from two users, and the channels are assumed to be fading. The problem is modeled by a discounted infinite horizon Markov decision process (MDP). The objective is to find…
New algorithm reduces offline RL data requirements significantly.
problem Optimizing policies using only historical data in reinforcement learning.
method Off-Policy Double Variance Reduction (OPDVR) algorithm.
result OPDVR achieves optimal sample complexity with O ( H 2 / d m ε 2 ) O(H^2/d_mε^2) O ( H 2 / d m ε 2 ) episodes. Improved convergence for actor-critic algorithms in MDPs.
problem Global convergence analysis for actor-critic algorithms in MDPs.
method Introduced an analytical framework to handle complex recursions, established convergence to ε-close globally optimal policy with improved sample complexity.
result Converges to ε-close globally optimal policy with sample complexity of O(ε^(-3)) compared to O(ε^(-2)) for ε-close stationary policy.
MVPI framework optimizes risk in reinforcement learning, improving performance in robot simulations.
problem Optimizing risk in reinforcement learning control problems.
method Mean-Variance Policy Iteration (MVPI) framework for risk-averse control in MDPs.
result Risk-averse TD3 outperforms previous methods in robot simulation tasks.
New algorithm LOOP learns infinite-horizon AMDPs efficiently with function approximation.
problem Learning optimal policies in infinite-horizon AMDPs with function approximation.
method LOOP combines model-based and value-based methods with novel confidence sets and policy updating.
result LOOP achieves sublinear regret bound of i l d e O ( p o l y ( d , s p ( V ∗ ) ) T β ) ilde{\mathcal{O}}(\mathrm{poly}(d, \mathrm{sp}(V^*)) \sqrt{Tβ} ) i l d e O ( poly ( d , sp ( V ∗ )) T β ) . New RL method learns K-step lookahead Q-functions for fixed-horizon MDPs.
problem Challenges in online reinforcement learning for non-episodic, finite-horizon MDPs.
method Introduces a K-step lookahead Q-function with a time-varying threshold for selecting actions.
result Achieves minimax optimal constant regret for K=1 and O ( max ( ( K − 1 ) , C K − 1 ) S A T log ( T ) ) \mathcal{O}(\max((K-1),C_{K-1})\sqrt{SAT\log(T)}) O ( max (( K − 1 ) , C K − 1 ) S A T log ( T ) ) regret for K ≥ 2. New algorithm solves constrained MDPs with generative model, reducing sample complexity.
problem Maximizing rewards while adhering to constraints in MDPs.
method Primal-dual framework using mirror descent value iteration.
result Near-optimal sample complexity bounds for CMDPs.