Optimistic NPG improves policy optimization in online RL with efficient sample complexity.
problem Limited theoretical understanding of policy optimization, especially in online RL.
method Combines natural policy gradient with optimistic policy evaluation.
result Achieves optimal dimension dependence sample complexity for learning near-optimal policies.
New algorithm reduces regret in sequential decision-making problems.
problem Balancing exploration and exploitation in online sequential decision problems.
method Variational Bayesian optimistic sampling (VBOS) for optimizing policies.
result VBOS achieves i l d e O ( A T ) ilde O(\sqrt{AT}) i l d e O ( A T ) Bayesian regret for stochastic multi-armed bandits. Accelerates policy optimization in RL with optimistic and adaptive updates.
problem Improving policy optimization methods in reinforcement learning.
method Integrates foresight into policy improvement step via optimistic and adaptive updates.
result Designs an optimistic policy gradient algorithm, adaptive via meta-gradient learning.
Develops ODRPO to improve RL algorithms with better performance and stability.
problem RL algorithms converge to sub-optimal solutions due to limited policy representation.
method Integrates DRO approach to solve trust region constrained optimization problem without parameterizing policies.
result Achieves globally optimal policy update and higher sample efficiency.
In this paper, we prove some convergence results of a special case of optimistic policy iteration algorithm for stochastic shortest path problem. We consider both Monte Carlo and T D ( λ ) TD(λ) T D ( λ ) methods for the policy evaluation step under the condition that the termination state will eventually be reached almost surely.
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.
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. PROPO tackles non-stationary MDPs with efficient policy optimization.
problem Non-stationary MDPs with varying reward and transition kernels.
method PROPO, a periodic restarted optimistic policy optimization algorithm with sliding-window-based policy evaluation and improvement.
result PROPO achieves near-optimal performance in non-stationary MDPs.
Proposes H-UCRL for efficient model-based RL with sublinear regret.
problem Greedy policy exploration in model-based RL ignores epistemic uncertainty.
method Reparameterizes plausible models, hallucinates control, augments input space, solves with greedy planners.
result H-UCRL achieves provably sublinear regret for Gaussian Process models.
New RL algorithm optimizes policies with bandit feedback, matching previous bounds.
problem Optimizing policies with unknown transitions and bandit feedback.
method Optimistic Trust Region Policy Optimization (TRPO) algorithm.
result Sub-linear regret bounds for both stochastic and adversarial rewards.
New method minimizes regret in AMDP with high probability.
problem Pessimistic regret analysis in AMDP.
method Developed a new policy search method achieving optimistic regret.
result Achieved sublinear optimistic regret with high probability.
While policy-based reinforcement learning (RL) achieves tremendous successes in practice, it is significantly less understood in theory, especially compared with value-based RL. In particular, it remains elusive how to design a provably efficient policy optimization algorithm that incorporates exploration. To bridge su…
New RL algorithm tackles online robust MDPs with uncertainty.
problem Developing robust reinforcement learning models for real-world environments.
method Proposes a robust optimistic policy optimization algorithm for online robust MDPs.
result Establishes the first regret bound for online robust MDPs.
A new policy for contextual bandits adapts to reward vector shifts.
problem Learning under reward vector shifts with ordered rewards.
method Adaptive-discretization and optimistic elimination policy.
result Established upper bounds on preference-based regret.
New batch RL method avoids overly optimistic policies.
problem Challenges in applying RL to large domains with limited data.
method Modified Bellman optimality and evaluation back-up for conservative updates.
result Can find approximately best policies within explored state-action space.
New algorithm achieves data-dependent regret bounds in MDPs with unknown transitions.
problem Achieving best-of-both-worlds guarantees with data-dependent regret bounds in MDPs with unknown transitions.
method Optimistic follow-the-regularized-leader algorithm with new optimistic Q-function estimators and transition bonus.
result First-order, second-order, and path-length bounds with polylog(T) regret in the stochastic regime.
DORIS algorithm achieves no-regret learning in Markov games with adversarial opponents.
problem Decentralized policy learning in Markov games with nonstationary opponents.
method DORIS algorithm using optimistic hyperpolicy mirror descent.
result Achieves K \sqrt{K} K -regret in general function approximation. 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.
Study optimal policy regret in partially observable Markov games with adaptive opponents.
problem Optimal sequential decision-making in partially observable environments against strategic, adaptive opponents.
method An epoch-based optimistic maximum-likelihood algorithm that selects one policy per epoch using confidence sets built cumulatively from past data.
result Achieves i l d e O ( T ) ilde{O}(\sqrt{T}) i l d e O ( T ) policy regret for fixed problem parameters, with explicit dependence on horizon, adversary memory, confidence radius, and aggregate Eluder dimension. Algorithm converges to Nash equilibria in competitive games.
problem Finding Nash equilibria in decentralized, competitive Markov games.
method Decentralized Optimistic Gradient Descent/Ascent with a critic.
result Converges to the set of Nash equilibria under self-play.
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.
Off-policy reinforcement learning (RL) using a fixed offline dataset of logged interactions is an important consideration in real world applications. This paper studies offline RL using the DQN replay dataset comprising the entire replay experience of a DQN agent on 60 Atari 2600 games. We demonstrate that recent off-p…
Improved POMDP regret to sqrt(T) with known observation model.
problem Average-reward POMDPs with unknown transition model but known observation model.
method Optimistic algorithm using deterministic policies and novel estimation techniques.
result First approach with regret guarantee of sqrt(T) against optimal policy.
We consider the problem of online planning in a Markov Decision Process when given only access to a generative model, restricted to open-loop policies - i.e. sequences of actions - and under budget constraint. In this setting, the Open-Loop Optimistic Planning (OLOP) algorithm enjoys good theoretical guarantees but is …
New algorithms reduce regret in online MDPs by adapting to data and variance.
problem Adapting to both adversarial and stochastic environments in online MDPs.
method Develops algorithms based on global optimization and policy optimization, using optimistic follow-the-regularized-leader with log-barrier regularization.
result Achieves refined data-dependent and variance-dependent regret bounds.
Paper proposes an RL algorithm to ensure policy performance guarantees.
problem Lack of performance guarantees for RL policies compared to baselines.
method Online model-free algorithm that ensures conservative exploration.
result Regret bound of i l d e O ( T ) ilde{\mathcal{O}}(\sqrt{T}) i l d e O ( T ) for both discrete and continuous spaces. In many cases an intelligent agent may want to learn how to mimic a single observed demonstrated trajectory. In this work we consider how to perform such procedural learning from observation, which could help to enable agents to better use the enormous set of video data on observation sequences. Our approach exploits t…
New rule reduces exploration regret to logarithmic, improving bad episode handling.
problem Improving exploration regret in average reward MDPs.
method Replacing Doubling Trick with Vanishing Multiplicative rule in EVI-based algorithms.
result Regret is logarithmic under the new rule, significantly better than linear.
Paper optimizes GAIL for online and offline learning with linear approximations.
problem Imitation learning from expert demonstrations with linear function approximations.
method Proposes optimistic and pessimistic algorithms for online and offline settings.
result Proves optimality and efficiency of proposed algorithms.
Robust Optimization has traditionally taken a pessimistic, or worst-case viewpoint of uncertainty which is motivated by a desire to find sets of optimal policies that maintain feasibility under a variety of operating conditions. In this paper, we explore an optimistic, or best-case view of uncertainty and show that it …
RANDomized-exploration policy Optimization via Multiple Importance Sampling with Truncation (RANDOMIST) for PO with mediator feedback.
problem Policy Optimization in continuous control tasks.
method RANDomized-exploration policy Optimization via Multiple Importance Sampling with Truncation (RANDOMIST) for regret minimization in PO.
result Achieving constant regret under certain circumstances in PO with mediator feedback.
Paper proposes a framework for reliable off-policy evaluation in reinforcement learning.
problem Quantifying uncertainty in off-policy estimates for safe deployment of target policies.
method Distributionally robust optimization for creating confidence bounds.
result Non-asymptotic and asymptotic guarantees for robust cumulative reward estimates.
Online Apprenticeship Learning aims to match expert performance without access to cost functions.
problem Learning an agent's policy to match expert performance in an MDP without cost function access.
method Combines mirror descent based no-regret algorithms for policy optimization and cost learning, with optimistic exploration.
result Derives an algorithm with O ( K ) O(\sqrt{K}) O ( K ) regret, practical for high-dimensional control problems. Optimistic algorithm reduces regret in non-stationary linear MDPs.
problem Efficient learning in non-stationary linear MDPs with evolving reward and transition.
method OPT-WLSVI, an optimistic model-free algorithm using exponential weights.
result Achieves a regret bound of O ~ ( d 5 / 4 H 2 Δ 1 / 4 K 3 / 4 ) \widetilde{\mathcal{O}}(d^{5/4}H^2 Δ^{1/4} K^{3/4}) O ( d 5/4 H 2 Δ 1/4 K 3/4 ) . Paper proposes a new dynamic pricing method with always-valid online statistical learning.
problem Designing dynamic pricing policies that adapt to online uncertainty and maintain validity.
method Regularized online statistical learning with theoretical guarantees and three major advantages.
result Proposed OORMLP pricing policy secures logarithmic regret in decision horizon.
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.
New algorithms ensure policies perform at least as good as a baseline in reinforcement learning.
problem Learning policies that are guaranteed to perform at least as well as a baseline in reinforcement learning.
method Introduce conservative exploration for average reward and finite horizon problems, presenting two optimistic algorithms.
result Guaranteed performance of policies at least as good as a baseline, without hindering learning ability.
Bayesian classifier improves robustness with optimistic score ratio.
problem Limited information on class-conditional distribution.
method Optimistic score ratio for robust binary classification.
result Bayesian classifier using optimistic score ratio is robust and computationally tractable.
The paper analyzes CMDPs, balancing exploration and exploitation to avoid constraint violations.
problem Balancing exploration and exploitation in CMDPs to satisfy constraints.
method Two approaches: optimistic planning and incremental updates of primal and dual variables.
result Both approaches achieve sublinear regret on utility and constraint violations, with stronger guarantees for the linear programming approach.
We show that the optimistic limits of the colored Jones polynomials of the hyperbolic knots coincide with the optimistic limits of the Kashaev invariants modulo 4 π 2 4π^2 4 π 2 .
Proposes Optimistic Pessimistically Initialised Q-Learning (OPIQ) for better exploration in RL.
problem Pessimistic initialisation of Q-values in deep RL leads to poor exploration performance.
method Augments pessimistically initialised Q-values with count-based bonuses to ensure optimism.
result OPIQ outperforms non-optimistic DQN variants in hard exploration tasks.
New method selects best offline RL policies from logged data.
problem Hyperparameter selection challenges offline RL.
method Offline hyperparameter selection for RL algorithms.
result Reliable ranking and selection of policies across hyperparameters.
We discuss the relative merits of optimistic and randomized approaches to exploration in reinforcement learning. Optimistic approaches presented in the literature apply an optimistic boost to the value estimate at each state-action pair and select actions that are greedy with respect to the resulting optimistic value f…
New algorithm learns optimal policy with multi-step lookahead information.
problem Learning optimal policy in reinforcement learning with multi-step lookahead information is NP-hard.
method Adaptive batching policies that process lookahead in state-dependent chunks.
result Order-optimal regret bounds up to a constant factor of lookahead horizon.
The optimistic limit is the mathematical formulation of the classical limit which is a physical method to expect the actual limit by using saddle point method of certain potential function. The original optimistic limit of the Kashaev invariant was formulated by Yokota, and a modified formulation was suggested by the a…
New principle optimizes bandit decisions with context.
problem Dealing with general function classes and large context spaces in bandits.
method Upper Counterfactual Confidence Bounds (UCCB) principle.
result Proves optimality and efficiency in complex settings.
Optimistic estimate predicts best fitting performance of nonlinear models.
problem Evaluating the potential of nonlinear models in fitting.
method Proposes an optimistic estimate to quantify the smallest sample size for fitting nonlinear models.
result Predicts specific subsets of targets that can be fitted at overparameterization.
New RL algorithm learns good actions from offline data, reducing uncertainty and divergence.
problem Limited applicability of current RL algorithms in real-world settings due to high costs of exploration.
method Proposes an algorithm for batch RL using a fixed offline dataset, with penalties for policy and value constraints.
result Compared favorably to state-of-the-art methods on 32 continuous-action benchmarks.