Paper establishes first instance-dependent lower bound for PAC reinforcement learning.
problem Identifying near-optimal policies in tabular MDPs with minimal samples.
method Proposes instance-dependent lower bound for sample complexity.
result Lower bound closely matches PEDEL algorithm's sample complexity.
UCRL-WVTR tackles long-term reinforcement learning with general approximations, achieving horizon-free and instance-dependent regret bounds.
problem Long-term reinforcement learning with general function approximations.
method UCRL-WVTR proposes a novel algorithm, UCRL-WVTR, with weighted value-targeted regression and a high-order moment estimator.
result Achieves horizon-free and instance-dependent regret bounds matching minimax lower bounds up to logarithmic factors.
New algorithm optimally evaluates policies with linear approximations.
problem Policy evaluation with linear function approximation.
method Accelerated, variance-reduced fast temporal difference algorithm (VRFTD).
result VRFTD matches both deterministic and stochastic lower bounds.
New methods improve temporal difference learning for policy evaluation in Markov decision processes.
problem Improving temporal difference learning for policy evaluation in Markov decision processes.
method Introduced variance-reduced forms of stochastic approximation to achieve non-asymptotic, instance-dependent optimality.
result Temporal difference learning is strictly suboptimal, but variance-reduced forms achieve optimality up to logarithmic factors.
Study optimal and instance-dependent guarantees for solving linear equations with Markovian data.
problem Approximately solving linear fixed point equations with Markovian data.
method Non-asymptotic bounds and instance-dependent characterizations for stochastic approximation.
result Instance-optimality of the averaged SA estimator and matching upper and lower bounds.
We consider PAC-learning a good item from k-subsetwise feedback information sampled from a Plackett-Luce probability model, with instance-dependent sample complexity performance. In the setting where subsets of a fixed size can be tested and top-ranked feedback is made available to the learner, we give an algorithm w…
Paper bounds PAC RL sample complexity in deterministic MDPs.
problem Identify ε-optimal policy with high probability.
method Proposes nearly matching upper and lower bounds on sample complexity, introduces deterministic return gap, uses graph-theoretical concepts and maximum-coverage exploration.
result First nearly matching upper and lower bounds on sample complexity for PAC RL in deterministic MDPs.
New algorithm tackles heavy-tailed rewards in RL with instance-dependent regret bounds.
problem Efficient algorithms for RL with heavy-tailed rewards in large state-action spaces.
method Design of \textsc{Heavy-OFUL} for heavy-tailed linear bandits and \textsc{Heavy-LSVI-UCB} for RL with linear function approximation.
result First instance-dependent regret bounds for heavy-tailed rewards in RL with linear function approximation.
The paper tackles a bandit problem with infinitely many arms per group, aiming to identify the group with the highest quantile reward.
problem Max-quantile group bandit problem with infinitely many arms per group.
method Two-step algorithm: first request arms from each group, then apply a finite-arm max-quantile bandit algorithm.
result Characterization of instance-dependent and worst-case regret, with matching lower bounds.
No communication allows optimal instance-dependent regret guarantees in multi-player bandits.
problem Achieving optimal instance-dependent regret in multi-player multi-armed bandits without communication.
method Characterization of Pareto optimal trade-offs and development of an algorithm.
result Achieving optimal instance-dependent regret requires strict sub-optimality in other regimes.
Flexible algorithms for maximizing rewards in structured bandits.
problem Reward maximization in structured stochastic multi-armed bandit problems.
method Asymptotically optimal algorithms using iterative saddle-point solvers.
result Achieves optimal performance with minimal computational burden.
Study optimizes solving fixed-point equations using subspace search.
problem Solving linear fixed point equations in Hilbert spaces.
method Linear stochastic approximation scheme with Polyak--Ruppert averaging.
result Established optimal approximation factor for temporal difference learning methods.
The study analyzes batched methods for early stopping in stochastic multi-armed bandits.
problem Early stopping in stochastic multi-armed bandits with fixed confidence.
method Instance-dependent lower bounds and a general batched algorithm with upper bounds.
result Upper and lower bounds on the number of batches and sample complexity.
An algorithm learns from multiple models to match an oracle's risk.
problem Learning from multiple noisy models to estimate a target parameter.
method Elimination rounds algorithm for adaptive learning.
result Risk of weak-oracle learner matches that of an oracle in multiple source case.
New study on regret lower bounds for multi-agent multi-armed bandit problems.
problem Understanding the limits of performance in multi-agent multi-armed bandit problems.
method Comprehensive study on different settings, establishing tight lower bounds.
result First comprehensive study on regret lower bounds across various settings.
VRPG algorithm optimizes convex constraints with non-asymptotic guarantees.
problem Stochastic convex optimization under convex constraints.
method Natural variance reduced proximal gradient (VRPG) algorithm.
result VRPG achieves local minimax lower bound up to constants and log factor of N. ERTS uses Thompson sampling for Gaussian entropic risk bandits, achieving regret bounds.
problem Risk in decision making complicates reward maximization in MAB problems.
method ERTS (Entropic Risk Thompson Sampling) using Thompson sampling with an entropic risk measure.
result Regret bounds for ERTS under entropic risk measure provided.
RL-LOW algorithm achieves exponential simple regret in offline RLHF with pairwise comparisons.
problem Offline reinforcement learning from human feedback with pairwise comparisons.
method Proposes RL-LOW algorithm to minimize simple regret with exponential convergence.
result Achieves an exponential form of simple regret of \(\exp ( - Ω(n/H) )\).
Optimal sequential testing for Markovian data with lower and upper bounds.
problem Sequential hypothesis testing for Markovian data.
method Non-asymptotic lower bounds and optimal test design.
result Optimal test matches lower bound asymptotically.
New algorithms improve contextual bandit performance by adapting to problem difficulty.
problem Improving contextual bandit performance on problems with varying difficulty.
method Introducing complexity measures and oracle-efficient algorithms.
result Achieves optimal instance-dependent regret bounds for rich policy classes.
Optimizes arm selection with side information in Gaussian bandits.
problem Optimizing arm selection with side information in Gaussian bandits.
method Constructs an LP-based asymptotic instance-dependent lower bound on the regret and develops the first known asymptotically optimal algorithm.
result First known asymptotically optimal algorithm for Gaussian bandits with side information.
This paper studies the statistical theory of batch data reinforcement learning with function approximation. Consider the off-policy evaluation problem, which is to estimate the cumulative value of a new target policy from logged history generated by unknown behavioral policies. We study a regression-based fitted Q iter…
New algorithm reduces regret for kernelized bandits by adapting to specific problem instances.
problem Efficiently learning the optimizer of an unknown function in RKHS with noisy oracle.
method Instance-dependent regret analysis and a new minimax near-optimal algorithm.
result New algorithm achieves better performance on specific problem instances.
New algorithm reduces regret by allowing free exploration in multi-armed bandits.
problem Designing an adaptive policy to minimize regret with a free exploration budget.
method Introduced (α,β)-probably saving policies and a two-phase algorithm UFE-KLUCB-H. result UFE-KLUCB-H accumulates strictly less regret than non-free exploration policies.
Optimal multi-fidelity best-arm identification reduces cost with better accuracy.
problem Finding the best arm with highest mean reward at minimum cost.
method Gradient-based approach with asymptotically optimal cost complexity.
result Asymptotically optimal cost complexity compared to existing methods.
A new SSL method uses instance-dependent thresholds to improve accuracy.
problem Improving semi-supervised learning by better selecting confident unlabeled instances.
method Proposes instance-dependent thresholds that vary based on the ambiguity and error rates of pseudo-labels for each unlabeled instance.
result Demonstrates that instance-dependent thresholds provide a probabilistic guarantee for correct pseudo-labels.
Study robust best-arm identification in linear bandits with lower bounds and algorithms.
problem Identify a near-optimal robust arm in linear bandits with adversarial actions.
method Propose instance-dependent lower bounds and both static and adaptive bandit algorithms.
result Sample complexity matches the lower bound and algorithms effectively identify robust arms.
COF algorithm minimizes cost in multi-armed bandits with known costs and reward constraints.
problem Minimizing cost while meeting a minimum reward requirement in uncertain environments.
method COF algorithm that intelligently combines samples from all arms to gauge feasibility and minimize cost.
result COF achieves instance-dependent upper bounds on cumulative cost and quality regret.
Improved stochastic optimization outperforms standard methods.
problem Optimizing smooth, strongly convex functions with noisy data.
method Variance reduction strategy called VISOR.
result VISOR achieves optimal sample complexity and oracle complexity.
New algorithm reduces contextual bandit identification to argmax calls.
problem Best-arm identification in stochastic contextual bandits.
method Instance-optimal PAC algorithm using argmax oracle calls.
result First instance-dependent PAC sample complexity for contextual bandits.
Two algorithms tackle heavy-tailed rewards in reinforcement learning with linear function approximation.
problem Online sequential decision-making with heavy-tailed rewards.
method AdaOFUL and VARA algorithms for linear stochastic bandits and MDPs, using modified adaptive Huber regression.
result Achieved state-of-the-art and variance-aware regret bounds for heavy-tailed rewards.
A new federated multi-armed bandit framework with personalization balances generalization and personalization.
problem Balancing generalization and personalization in federated multi-armed bandits.
method Proposed a Personalized Federated Upper Confidence Bound (PF-UCB) algorithm to achieve a O(log(T)) regret. result PF-UCB achieves an O(log(T)) regret regardless of personalization degree and has similar instance dependency to lower bound. The paper achieves nearly optimal regret bounds for contextual multinomial logit bandits.
problem The contextual multinomial logit (MNL) bandit problem with varying rewards.
method Established lower bounds and proposed OFU-MNL+ algorithm with matching upper bounds.
result Achieved minimax optimal regret bounds for both uniform and non-uniform reward settings.
Algorithm aggregates rewards from multiple players to learn related tasks in online bandit learning.
problem Learning related but slightly different tasks in an online setting with heterogeneous feedback.
method RobustAgg(ε) algorithm that aggregates rewards from different players. result Achieves instance-dependent regret guarantees and nearly matching lower bounds.
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.
Study on policy testing in MDPs with lower bounds and new algorithm.
problem Deciding if policy value exceeds a threshold with limited samples.
method Derived lower bound, proposed new algorithm, reformulated problem, used policy optimization in reversed MDP.
result New algorithm outperforms existing methods in policy testing.
Bandits with Knapsacks (BwK) is a general model for multi-armed bandits under supply/budget constraints. While worst-case regret bounds for BwK are well-understood, we present three results that go beyond the worst-case perspective. First, we provide upper and lower bounds which amount to a full characterization for lo…
New reinforcement learning algorithm achieves instance-optimal sample complexity.
problem Achieving low regret and identifying optimal policies in reinforcement learning.
method A novel planning-based algorithm that explicitly accounts for state visitation distributions.
result The proposed algorithm attains nearly minimax optimal sample complexity, improving over worst-case bounds.
Study quantile reward identification with 1-bit feedback constraints.
problem Best arm identification with quantile reward and 1-bit communication.
method Proposes an algorithm using noisy binary search for quantile reward estimation.
result Derives upper and lower bounds on sample complexity for 1-bit feedback.
New method estimates optimal Q-values with better accuracy for specific problems.
problem Estimating optimal Q-values in reinforcement learning is difficult and varies by problem instance.
method Local minimax framework and variance-reduced Q-learning.
result Sharp lower bounds on estimation accuracy for Q-learning.
Paper tackles instance-dependent label noise by approximating it with part-dependent noise.
problem Learning with instance-dependent label noise is challenging.
method Approximate instance-dependent label noise with part-dependent noise. Use transition matrices for parts to model noise.
result Method outperforms state-of-the-art approaches for instance-dependent label noise.
ACOL learns constraints from human preferences in driving simulations.
problem Learning constraints from human preferences in driving simulations.
method Adaptive Constraint Learning (ACOL) algorithm for constrained linear best-arm identification.
result ACOL's sample complexity matches worst-case lower bound and is significantly tighter in the average case.
New algorithm identifies best arm in semiparametric bandits with near optimal efficiency.
problem Fixed-confidence Best Arm Identification in semiparametric bandits with unknown baseline shift.
method Phase-elimination algorithm based on orthogonalized regression design.
result Nearly optimal high-probability sample-complexity upper bound established.
New bounds explain neural network generalization by considering local Lipschitz properties.
problem Existing bounds fail to account for initialization and SGD biases.
method Optimal transport interpretation of generalization problem.
result Instance-dependent bounds that depend on local Lipschitz regularity.
New CTRL algorithm adapts to varying problem difficulty.
problem Adapting to varying levels of problem difficulty in CTRL.
method MLE with a general function approximator, estimating state marginal density.
result Regret bound scales with reward variance and measurement resolution, independent of measurement strategy.
BeGIN benchmarks GNNs for instance-dependent label noise in graphs.
problem Instance-dependent label noise in graph data.
method BeGIN introduces a benchmark with various noise types and evaluates noise-handling strategies across GNN architectures.
result Challenges of instance-dependent noise, especially LLM-based corruption, and the importance of node-specific parameterization.
Stochastic linear bandits are a natural and simple generalisation of finite-armed bandits with numerous practical applications. Current approaches focus on generalising existing techniques for finite-armed bandits, notably the optimism principle and Thompson sampling. While prior work has mostly been in the worst-case …
This work improves the convergence theory of diffusion models for generating samples from complex distributions.
problem Improving theoretical understanding of diffusion models, particularly their convergence analysis.
method Developed an instance-dependent convergence rate that adapts to the smoothness of target distributions.
result Established an iteration complexity of min{d,d2/3L1/3,d1/3L}ε−2/3 for generating high-quality samples.