New algorithm identifies dominant arm with high probability.
problem Identifying the arm with the highest realized reward in multi-armed bandits.
method Dominance score criterion and joint mixing and recycling mechanism.
result Identifies the best dominant arm with nearly optimal sample complexity.
A new algorithm for competing agents in a two-sided market setting.
problem Decentralized competition between agents in a two-sided market with unknown valuations.
method UCB-D3 algorithm for UCB with Decentralized Dominant-arm Deletion.
result UCB-D3 is order optimal and achieves a new regret lower bound.
New study shows non-adaptive trials can be outperformed by adaptive designs in treatment selection.
problem Determining the best allocation of resources in clinical trials.
method Analysis of batched arm elimination designs and comparison with completely randomized trials.
result Simple adaptive designs universally and strictly dominate non-adaptive completely randomized trials for at least three treatment arms.
We study a strategic version of the multi-armed bandit problem, where each arm is an individual strategic agent and we, the principal, pull one arm each round. When pulled, the arm receives some private reward v a v_a v a and can choose an amount x a x_a x a to pass on to the principal (keeping v a − x a v_a-x_a v a − x a for itself). All non-pulle…
Optimal algorithms identify non-dominated arms in multi-output linear bandit models.
problem Identifying the Pareto Set in multi-output linear bandit models.
method Design-based algorithms for Pareto Set Identification (PSI) in a structured multi-output linear bandit model.
result Nearly optimal guarantees in both fixed-budget and fixed-confidence settings.
In the Best- k k k -Arm problem, we are given n n n stochastic bandit arms, each associated with an unknown reward distribution. We are required to identify the k k k arms with the largest means by taking as few samples as possible. In this paper, we make progress towards a complete characterization of the instance-wise sample…
The Knowledge Gradient (KG) policy was originally proposed for online ranking and selection problems but has recently been adapted for use in online decision making in general and multi-armed bandit problems (MABs) in particular. We study its use in a class of exponential family MABs and identify weaknesses, including …
This paper shows ARMs and EBMs are equivalent, revealing ARM lookahead capabilities.
problem Understanding the lookahead capabilities of next-token prediction models.
method Unified view of ARMs and EBMs, establishing a bijection and deriving equivalence.
result ARMs and EBMs are equivalent, revealing ARM lookahead capabilities.
Thompson Sampling tackles USS, a sequential selection problem without feedback.
problem Unsupervised Sequential Selection (USS) problem with fixed costs and ordered arms.
method Thompson Sampling algorithm for USS problem.
result Thompson Sampling achieves near optimal regret and better performance than existing algorithms.
New algorithm reduces bandit regret by graph domination number.
problem Adversarial multi-armed bandit with partial observations and switching costs.
method New algorithm with improved policy regret bounds.
result Regret depends only on the domination number of the feedback graph.
New graph feedback model for bandits with improved regret bounds.
problem Understanding how graph structure affects regret in bandit problems.
method Introduced fractional weak domination number and k k k -packing independence number to capture upper and lower bounds on regret. Used strong duality theorem to derive upper and lower bounds. result Proved general upper and lower bounds on regret for various graph structures, showing tightness up to a logarithmic factor.
We consider the partial observability model for multi-armed bandits, introduced by Mannor and Shamir. Our main result is a characterization of regret in the directed observability model in terms of the dominating and independence numbers of the observability graph. We also show that in the undirected case, the learner …
Optimal regret achieved in stochastic, discrete multi-armed bandits using information-theoretic exploration.
problem Optimal exploration vs. exploitation in stochastic, discrete multi-armed bandits.
method Proposes an information-theoretic strategy based on the value of information criterion, using simulated-annealing-like updates of a parameter.
result Achieves logarithmic optimal regret with respect to the number of episodes.
Algorithm learns optimal arm selection in unsupervised sequential selection with contextual information.
problem Learning optimal arm selection in unsupervised sequential selection with contextual information.
method Proposes an algorithm for the contextual USS problem under the CWD property, demonstrating sub-linear regret.
result Demonstrates sub-linear regret for the proposed algorithm.
New method assesses multivariate stochastic dominance using Optimal Transport.
problem Benchmarking models across multiple metrics considering dependencies.
method Characterization of multivariate first stochastic dominance via couplings, entropic regularization, and Optimal Transport.
result Established CLT and consistency for the empirical statistic, enabling hypothesis testing.
In this paper, we study the stochastic combinatorial multi-armed bandit (CMAB) framework that allows a general nonlinear reward function, whose expected value may not depend only on the means of the input random variables but possibly on the entire distributions of these variables. Our framework enables a much larger c…
New bandit model for e-commerce with ordered categories.
problem Optimizing customer experience in e-commerce with unknown preferences.
method Introducing three types of ordering between categories and proving lower bounds on cumulative regret.
result Proved that algorithms can fully leverage the structure of the model with theoretical guarantees.
Paper introduces Decentralized Non-stationary Competing Bandits ( exttt{DNCB}) for dynamic matching markets.
problem Understanding dynamic two-sided matching markets with competing agents.
method Proposes a decentralized asynchronous learning algorithm ( exttt{DNCB}) for non-stationary environments.
result Obtains sub-linear (logarithmic) regret of exttt{DNCB} in dynamic settings.
A new algorithm THV-UCB reduces regret in multi-objective bandit problems.
problem Maintaining a small set of actions that jointly approximate the Pareto frontier in multi-objective slate selection.
method THV-UCB, an optimistic algorithm that selects arms based on optimistic estimates of their marginal hypervolume contributions.
result The algorithm achieves a gap-free regret bound of i l d e O ( d n k T ) ilde{O}(d\sqrt{nkT}) i l d e O ( d nk T ) and a gap-dependent bound of i l d e O ( n k 2.5 / Δ min ) ilde{O}(nk^{2.5}/Δ_{\min}) i l d e O ( n k 2.5 / Δ m i n ) . New algorithms reduce matching market regret to log(T) with improved stability.
problem Minimizing regret in two-sided matching markets with bandit feedback.
method Phase-based algorithm with local arm deletion to improve stability.
result Achieves Θ(log(T)) regret for markets with uniqueness consistency.
The stochastic multi-armed bandit model is a simple abstraction that has proven useful in many different contexts in statistics and machine learning. Whereas the achievable limit in terms of regret minimization is now well known, our aim is to contribute to a better understanding of the performance in terms of identify…
This paper studies the Best-of-K Bandit game: At each time the player chooses a subset S among all N-choose-K possible options and observes reward max(X(i) : i in S) where X is a random vector drawn from a joint distribution. The objective is to identify the subset that achieves the highest expected reward with high pr…
Matching Markets meet Cumulative Prospect Theory: Towards Optimal and Adversarially Robust Learning
problem Multi-agent multi-armed bandit problem in competitive setup with two-sided matching markets under human-centric decision making model
method Using cumulative prospect theory (CPT) to emulate human preferences
result Improved regret guarantees in adversarial markets with CPT as risk-sensitive measure
This paper unifies risk-averse Thompson sampling for continuous risk functionals.
problem Designing and analyzing risk-averse Thompson sampling algorithms for continuous risk functionals.
method Developed analytical toolkits to prove asymptotically optimal regret bounds for various risk measures.
result Proved asymptotic optimality of ρ ρ ρ -MTS for Bernoulli distributions and a class of risk measures. Study uses few-shot learning to analyze claims and arguments in German debate on arms deliveries.
problem Limited data and computational resources for automated content analysis.
method Multilingual transformer model with adapter extension and few-shot learning.
result Parameter-efficient approach performs well on varying training set sizes.
Factored bandits model learns with limited feedback using decomposable actions.
problem Limited feedback learning with decomposable actions.
method Introduces factored bandits model, provides anytime algorithm, and matching upper and lower bounds.
result Improves regret bounds for utility-based dueling bandits.
Study on a bad arm existence checking problem to minimize arm draws.
problem Judging the existence of a positive arm among K given arms.
method Proposes an algorithm with arm selection policy and stopping condition.
result Proves the effectiveness of the proposed algorithm theoretically and empirically.
Adaptive framework improves airline pricing models' performance.
problem No single model dominates other models for all customer requests.
method Adaptive meta-decision framework using Thompson sampling.
result Improves expected revenue per offer by 43% and conversion score by 58%.
New algorithms for streaming bandits with limited memory.
problem Optimizing decisions in a stream of uncertain outcomes with limited memory.
method Developed algorithms for minimizing regret and identifying the best arm under bounded memory constraints.
result Upper and lower bounds on sample complexity for best-arm identification algorithms.
Study best arm identification in restless Markov multi-armed bandits with state-dependent transitions.
problem Identify the best arm in a multi-armed bandit with time-varying states.
method Propose a sequential policy to select arms without knowing their exact TPMs.
result Upper and lower bounds on expected time to find the best arm match in a special case.
Paper tackles identifying an odd arm in a multi-armed bandit with restless Markov processes and trembling hand.
problem Identifying an odd arm in a multi-armed bandit with restless Markov processes and trembling hand.
method Derive asymptotic lower bound on expected time to identify the odd arm, stitch together parameterised solutions to MDPs.
result First known asymptotic lower bound on expected time to identify the odd arm, with vanishing error probability.
A new bandit framework reduces K-armed to C+1-armed, achieving lower regret.
problem Designing efficient algorithms for correlated multi-armed bandits.
method Generalized UCB algorithm exploiting latent random source correlation.
result Achieves O ( 1 ) \mathcal{O}(1) O ( 1 ) regret for certain regimes, reducing from logarithmic. Algorithm identifies top M arms from K in stochastic bandits with limited budget.
problem Identifying the top M arms from K in a stochastic bandit setting with limited exploration budget.
method Develops an iterative algorithm that allocates budget nonlinearly to deactivate arms, deciding acceptance or rejection based on a decision rule.
result The algorithm effectively identifies the top M arms with a decay rate of misidentification probability characterized by nonlinear budget allocation.
Improved theoretical guarantees for Top Two algorithms.
problem Theoretical support for best arm identification with bounded distributions.
method General analysis of Top Two methods, identifying desirable properties and replacing sampling step.
result Theoretical support for Top Two algorithms with bounded distributions.
Sequential screening and dynamic regret in multi-armed bandits with arriving arms
problem Sequential experimentation with expanding arm set
method UCB-AA with preliminary screening
result Regret bounds depend on arrival process
Paper tackles good arm identification in stochastic bandits.
problem Identifying good arms with minimal samples.
method Proposes DGAI, a differentiable algorithm to improve sample complexity.
result DGAI outperforms baseline algorithms in synthetic and real-world datasets.
We consider the best-arm identification problem in multi-armed bandits, which focuses purely on exploration. A player is given a fixed budget to explore a finite set of arms, and the rewards of each arm are drawn independently from a fixed, unknown distribution. The player aims to identify the arm with the largest expe…
A novel algorithm reduces communication costs in federated best arm identification.
problem Identifying the best arm in a federated multi-armed bandit setup with minimal communication cost.
method Proposes a novel algorithm called FedElim that communicates only in exponential time steps.
result Demonstrates that communication is almost cost-free in FedElim, with a total cost at most 3 times the maximum under its variant.
New algorithms identify best arm with less pulls, adapting to arm covariances.
problem Best arm identification under dependent and correlated arm distributions.
method Adaptive algorithms estimating arm covariances to minimize pulls.
result Substantial improvement in best arm identification over standard setting.
Study best arm identification in restless bandits with unknown TPMs.
problem Identify the best arm with fixed confidence in restless bandits with unknown TPMs.
method Proposed a policy for best arm identification and proved its expected stopping time matches the lower bound.
result The state-action visitation proportions match the optimal proportions under any asymptotically optimal policy.
New method for identifying best arm in batched multi-armed bandit problems.
problem Identifying the best arm in multi-armed bandit problems where arms are sampled in batches.
method General linear programming framework for best arm identification in batched multi-armed bandit problems.
result Demonstrated good performance in numerical studies compared to UCB-type or Thompson sampling methods.
Optimal best-arm identification with known number of optimal arms.
problem Identifying the best arm in a multi-armed bandit with multiple optimal arms under fixed confidence.
method Deriving a new information-theoretic lower bound and proposing a modified stopping rule.
result Achieving asymptotic instance-optimality with a new lower bound and new stopping rule.
Paper tackles outlier detection in multi-armed bandits, achieving high accuracy with reduced exploration costs.
problem Detecting outlier arms in multi-armed bandit settings.
method Proposes GOLD algorithm based on upper confidence bounds to identify generic outlier arms.
result Achieves 98% accuracy with 83% reduction in exploration cost compared to state-of-the-art techniques.
CTS reduces regret in probabilistically triggered combinatorial bandits.
problem Optimizing decisions with probabilistically triggered arms in combinatorial multi-armed bandits.
method Combinatorial Thompson Sampling (CTS) with a regret bound analysis.
result Derives an O ( ∑ i = 1 m log T / ( p i Δ i ) ) O(\sum_{i =1}^m \log T / (p_i Δ_i)) O ( ∑ i = 1 m log T / ( p i Δ i )) regret bound for CTS. Improved best-arm identification in correlated multi-armed bandits.
problem Best-arm identification in multi-armed bandits with correlated rewards.
method Proposed C-LUCB algorithm that exploits upper bounds on conditional rewards.
result Significant reduction in sample complexity for best-arm identification.
Optimal best-arm identification in linear bandits reduces sampling budget.
problem Identifying the best arm with fixed confidence in stochastic linear bandits.
method A simple algorithm that tracks an optimal proportion of arm draws, updated as rarely as desired.
result The algorithm's sampling complexity matches known lower bounds, asymptotically almost surely and in expectation.
Algorithm reduces regret in restless multi-armed bandits by adaptively sequencing arm choices.
problem Minimizing regret in restless multi-armed bandits with unknown dynamics.
method Adaptive Sequencing Rules (ASR) algorithm that selects arms in a consecutive manner.
result Achieves logarithmic regret order with time and finite-sample bound.
New algorithm for identifying best drug arm in generalized linear bandits.
problem Identifying the best drug arm in drug design with minimal trials.
method Proposes an algorithm for best-arm identification in generalized linear bandits, providing theoretical guarantees and simulations.
result First algorithm for best-arm identification in generalized linear bandits with theoretical guarantees.