Policy gradient methods achieve linear convergence in simple MDPs.
problem Analyzing convergence rates of policy gradient methods in finite MDPs.
method Connections with policy iteration to show linear convergence with large step-sizes.
result Policy gradient methods succeed with large step-sizes and achieve linear rate of convergence.
Novel framework for policy optimization with general parameterization and linear convergence.
problem Lack of theoretical guarantees for policy optimization with general parameterization schemes.
method Mirror descent approach for policy optimization with general parameterization.
result First result of linear convergence for policy-gradient-based method with general parameterization.
This paper studies a class of continuous-time scalar-state stochastic Linear-Quadratic (LQ) optimal control problem with the linear control constraints. Applying the state separation theorem induced from its special structure, we develop the explicit solution for this class of problem. The revealed optimal control poli…
New algorithm for model selection in contextual bandits reduces regret.
problem Adapting to the complexity of the optimal policy in contextual bandits.
method Designing an algorithm that balances exploration and exploitation, achieving optimal regret bounds.
result Achieves i l d e O ( T 2 / 3 d m ⋆ 1 / 3 ) ilde{O}(T^{2/3}d^{1/3}_{m^\star}) i l d e O ( T 2/3 d m ⋆ 1/3 ) regret with no prior knowledge of the optimal dimension d m ⋆ d_{m^\star} d m ⋆ . Proposes a convergent TD algorithm for off-policy RL.
problem Learning value function from different policies in RL.
method Convergent on-policy TD algorithm with linear function approximation.
result Proposes a convergent TD algorithm for off-policy RL.
Study model selection in batch policy optimization with three error sources.
problem Learn a policy competitive with the best model class in batch policy optimization.
method Formalized in contextual bandit setting with linear model classes, addressing approximation error, statistical complexity, and dataset shift.
result No algorithm can optimally trade-off all three error sources, but relaxing any one enables near-oracle inequalities for the others.
ENIAC method optimizes and explores complex RL problems with non-linear policies.
problem Theoretical understanding of non-linear policies in RL with strategic exploration.
method ENIAC, an actor-critic method for non-linear function approximation.
result ENIAC finds near-optimal policies in polynomial exploration rounds under bounded eluder dimension.
Study online control of unknown time-varying systems with negative and positive results.
problem Online control of time-varying systems with unknown dynamics.
method Algorithmic upper bounds and lower bounds for different policy classes.
result Sublinear adaptive regret bounds for Disturbance Response policies.
We study derivative-free methods for policy optimization over the class of linear policies. We focus on characterizing the convergence rate of these methods when applied to linear-quadratic systems, and study various settings of driving noise and reward feedback. We show that these methods provably converge to within a…
Improved algorithms solve multi-period multi-class packing problems with bandit feedback.
problem Optimizing item packing under budget constraints with class-dependent rewards and bandit feedback.
method Developed a new estimator and a closed-form bandit policy for linear contextual multi-class multi-period packing problems.
result The proposed policy achieves sublinear regret in non-degenerate contexts, significantly outperforming benchmarks.
GPMD solves regularized RL with linear convergence, promoting structural policies.
problem Regularized reinforcement learning to encourage exploration and structural policies.
method Policy mirror descent with generalized convex regularizers and Bregman divergence.
result GPMD converges linearly to the global solution over a wide range of learning rates.
We study in this paper a class of constrained linear-quadratic (LQ) optimal control problem formulations for the scalar-state stochastic system with multiplicative noise, which has various applications, especially in the financial risk management. The linear constraint on both the control and state variables considered…
This work characterizes conditions for offline policy evaluation in reinforcement learning.
problem Understanding when classical methods succeed in offline policy evaluation for linear function approximation.
method Control-theoretic and linear-algebraic conditions for classical methods (FQI and LSTD).
result A precise hierarchy of regimes under which these estimators succeed, and a complete picture of their behavior.
This work analyzes policy gradient methods in reinforcement learning, providing convergence and approximation guarantees.
problem Theoretical convergence and approximation error of policy gradient methods in reinforcement learning.
method Analysis of policy gradient methods in discounted MDPs, focusing on tabular and parametric policy classes.
result Provably characterizations of computational, approximation, and sample size properties of policy gradient methods.
This paper studies GAIL's global convergence for general MDP and nonlinear rewards.
problem Understanding when GAIL algorithms achieve global convergence for general MDP and nonlinear rewards.
method Characterization of global convergence for various policy gradient algorithms applied to GAIL.
result First systematic theoretical study of GAIL for global convergence.
DMQ learns near-optimal policies efficiently with linear approximations.
problem Efficiently learning near-optimal policies with function approximation in reinforcement learning.
method DMQ algorithm with DSEC oracle for linear function approximation.
result DMQ returns a near-optimal policy using polynomial trajectories under certain assumptions.
Policy gradient methods converge for LQR problems with noisy state dynamics.
problem Finding optimal policies in noisy LQR problems over finite time horizons.
method Policy gradient methods with convergence guarantees for finite time and stochastic state dynamics.
result Global linear convergence for policy gradient methods in LQR problems with weak assumptions.
New approach generates optimal disturbances for controller verification.
problem Optimizing disturbances for controller verification with blackbox access.
method Online learning approach that adaptively generates disturbances based on controller inputs.
result New algorithm (MOTR) outperforms existing methods in simulated examples.
Optimistic PPO variant solves linear MDPs with improved regret bound.
problem Understanding theoretical limits of PPO in linear MDPs.
method Proposes an optimistic variant of PPO for episodic adversarial linear MDPs with full-information feedback.
result Establishes a i l d e O ( d 3 / 4 H 2 K 3 / 4 ) ilde{\mathcal{O}}(d^{3/4}H^2K^{3/4}) i l d e O ( d 3/4 H 2 K 3/4 ) regret bound. New RL method learns to skip states in linearly q π q^π q π -realizable MDPs, simplifying to linear MDPs.
problem Online RL in episodic MDPs with linearly q π q^π q π -realizable action-values. method Derives a novel algorithm that learns to skip states and applies a linear MDP algorithm.
result First polynomial-sample-complexity online RL algorithm for linearly q π q^π q π -realizable MDPs. Efficient RL algorithm for MDPs with linear Q π Q^π Q π realizability, achieving optimal regret bound.
problem Efficient reinforcement learning under linear Q π Q^π Q π realizability assumption for MDPs with stochastic dynamics. method Frozen Policy Iteration algorithm that uses high-confidence data and freezes policy for well-explored states.
result Achieves optimal regret bound of O ~ ( d 2 H 6 T ) \widetilde{O}(\sqrt{d^2H^6T}) O ( d 2 H 6 T ) for linear (contextual) bandits. DR-RPO optimizes robust policies in RL with limited interaction, achieving sublinear regret.
problem Policy optimization in RL under distribution shift and adversarial dynamics.
method DR-RPO algorithm incorporating reference-policy regularization and upper confidence bonus for exploration.
result DR-RPO achieves sublinear regret and polynomial suboptimality bounds in robust RL.
Paper formulates mutual information optimal control for discrete-time systems.
problem Optimal control of discrete-time linear systems with mutual information.
method Formulates MIOCP as an extension of MEOCP, derives optimal policy and prior, proposes alternating minimization algorithm.
result Proposes an alternating minimization algorithm for MIOCP.
New algorithms solve robust MDPs efficiently, significantly faster than existing methods.
problem Computing robust MDP solutions with uncertainty in transition probabilities is computationally expensive.
method Partial policy iteration and fast robust Bellman operator computation methods.
result The proposed methods are many orders of magnitude faster than state-of-the-art approaches.
BCRL learns a Bellman complete representation for offline RL policy evaluation.
problem Learning a Q-function efficiently from offline data.
method BCRL learns a linear Bellman complete representation directly from data, enabling efficient OPE.
result BCRL achieves competitive OPE error and outperforms FQE in certain scenarios.
New trading policies preserve robust gains in presence of transaction costs.
problem Maintaining robust gains in asset trading with transaction costs.
method Proposed double linear trading policies, analyzed with Monte Carlo simulations and historical data.
result Desired robust positive expected gain can be preserved under certain conditions.
Paper proves convergence of actor-critic methods for MDPs.
problem Finding approximate solutions to entropy-regularized MDPs.
method Stochastic approximation and policy evaluation with general regularizers.
result Probability one convergence of actor-critic algorithms.
We study the problem of policy evaluation and learning from batched contextual bandit data when treatments are continuous, going beyond previous work on discrete treatments. Previous work for discrete treatment/action spaces focuses on inverse probability weighting (IPW) and doubly robust (DR) methods that use a reject…
We address the problem of finding an optimal policy in a Markov decision process under a restricted policy class defined by the convex hull of a set of base policies. This problem is of great interest in applications in which a number of reasonably good (or safe) policies are already known and we are only interested in…
New framework learns policies for partially observable systems.
problem Learning policies in partially observable dynamical systems.
method Partially Observable Bilinear Actor-Critic framework.
result Algorithm can learn against optimal policies in certain cases.
This paper optimizes off-policy evaluation in reinforcement learning with function approximation.
problem Estimating cumulative value of a new policy from logged data generated by an unknown policy.
method Regression-based fitted Q iteration method, equivalent to estimating conditional mean embedding of transition operator.
result The method is minimax-optimal, with nearly minimal estimation error.
Papers learn from data to make decisions without interacting, improving on previous methods.
problem Achieving optimal decision-making from offline data with non-linear function approximation.
method Pessimistic Nonlinear Least-Square Value Iteration (PNLSVI) with three innovative components.
result Achieves minimax optimal instance-dependent regret for non-linear function approximation.
The paper analyzes the sample complexities for policy evaluation with linear function approximation.
problem Policy evaluation with linear function approximation in discounted infinite horizon Markov decision processes.
method Investigates sample complexities for two policy evaluation algorithms: TD and TDC.
result Establishes high-probability sample complexity bounds for policy evaluation algorithms.
Policy gradient converges to globally optimal policy in nearly linear-quadratic systems.
problem Finding optimal policies in nonlinear control systems with partial information.
method Policy gradient algorithm designed for nearly linear-quadratic regulators with small Lipschitz nonlinear components.
result Policy gradient algorithm converges to globally optimal policy with linear rate.
Q-MMR evaluates policies using reweighted rewards and moment matching.
problem Off-policy evaluation in finite-horizon MDPs.
method Q-MMR learns scalar weights for data points via a moment matching objective against a value-function discriminator class.
result Data-dependent finite-sample guarantee with a dimension-free error bound.
The paper provides a non-asymptotic error bound for linear system identification under nonlinear policies.
problem System identification for linear systems with nonlinear and/or time-varying policies under i.i.d. random excitation noises.
method Least square estimation with non-asymptotic error bound for bounded state and action trajectories.
result The error bound is consistent with linear policies and generalizes existing guarantees.
The paper proposes a policy learning framework for interpretable personalization.
problem Effective personalization of goods and services to improve revenues and maintain competitive edge.
method Policy learning with linear decision boundaries using causal inference and Bayesian optimization.
result The learned policy improves net sales revenue by 88.2% and provides insights into important features.
Presents SPEED, an algorithm for optimal policy evaluation in linear bandits with heteroscedastic noise.
problem Optimal data collection for policy evaluation in linear bandits with heteroscedastic reward noise.
method Formulated an optimal design for weighted least squares estimates, derived the optimal sample allocation, introduced SPEED algorithm, and derived regret bounds.
result SPEED leads to policy evaluation with MSE comparable to oracle strategy and significantly lower than random policy execution.
New algorithm tackles dynamic query routing to multiple embedding models.
problem Dynamic query routing to multiple embedding models under adversarial conditions.
method Formalized as adversarial contextual linear bandit with low-rank experts, proposed HPG algorithm.
result HPG algorithm achieves linearized policy regret of i l d e O ( s M T ) ilde{\mathcal O}(s\sqrt{M T}) i l d e O ( s M T ) . Enhances PMD with lookahead to improve RL performance.
problem Improving RL performance with greedy policies over 1-step.
method Integrates multi-step greedy policies into PMD with lookahead.
result Shows faster convergence rate for h h h -PMD. New algorithm learns policies without explicit rewards for MDPs.
problem Learning policies for MDPs without explicit rewards.
method UCRL-RFE algorithm with linear function approximation.
result UCRL-RFE achieves optimal policy with minimal exploration samples.
Study shows TD(0) with linear approx. converges for reversible Markov chains.
problem TD(0) divergence with off-policy and function approximation.
method Analyzes standard TD(0) with reversible Markov chains, adapting stochastic approximation framework.
result Establishes convergence with probability one for projected Bellman error = 0.
Optimistic actor-critic tackles linear MDPs with parametric policies.
problem Theoretical limitations of existing actor-critic methods for linear MDPs.
method Proposes an optimistic actor-critic framework with parametric log-linear policies and approximate Thompson sampling.
result Achieves state-of-the-art sample complexity in both on-policy and off-policy settings.
Paper proposes FPG algorithm for unbiased off-policy PG estimation.
problem Challenges in off-policy policy gradient estimation.
method Double Fitted PG estimation (FPG) algorithm for arbitrary policy parameterization.
result Empirically, FPG significantly outperforms existing methods.
Designs a single policy for collecting data to train near-optimal policies.
problem Engineering overhead in deploying minimax procedures for stochastic linear contextual bandits.
method Designs a single stochastic policy to collect data from which a near-optimal policy can be extracted.
result The designed policy can collect data from which a near-optimal policy can be extracted.
New RL algorithm achieves nearly optimal performance for linear MDPs.
problem Optimal reinforcement learning for episodic linear MDPs.
method Weighted linear regression with variance estimator and rare-switching policy.
result Achieves nearly minimax optimal regret i l d e O ( d H 3 K ) ilde O(d\sqrt{H^3K}) i l d e O ( d H 3 K ) . In this paper we study a model-based approach to calculating approximately optimal policies in Markovian Decision Processes. In particular, we derive novel bounds on the loss of using a policy derived from a factored linear model, a class of models which generalize numerous previous models out of those that come with s…
Demonstration-regularized RL reduces sample complexity for policy identification.
problem Improving reinforcement learning's sample efficiency with expert demonstrations.
method KL-regularization of behavior cloning using expert demonstrations.
result Demonstration-regularized RL achieves optimal policy identification with reduced sample complexity.