![[2602.14580] Replicable Constrained Bandits](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14580] Replicable Constrained Bandits
Summary
The paper discusses replicable online learning algorithms in constrained multi-armed bandit (MAB) problems, demonstrating that replicability can be achieved without sacrificing performance in terms of regret and constraint violation.
Why It Matters
This research addresses the critical need for reproducibility in machine learning experiments, particularly in constrained environments. By establishing replicable algorithms, it enhances the reliability of machine learning applications, which is essential for both academic research and practical implementations.
Key Takeaways
- Introduces the concept of replicability in constrained MAB problems.
- Demonstrates that replicable algorithms can match the performance of non-replicable ones.
- Develops the first replicable UCB-like algorithm for unconstrained MABs.
- Highlights the importance of optimism in-the-face-of-uncertainty for replicability.
- Contributes to the broader discourse on reproducibility in machine learning.
Computer Science > Machine Learning arXiv:2602.14580 (cs) [Submitted on 16 Feb 2026] Title:Replicable Constrained Bandits Authors:Matteo Bollini, Gianmarco Genalti, Francesco Emanuele Stradi, Matteo Castiglioni, Alberto Marchesi View a PDF of the paper titled Replicable Constrained Bandits, by Matteo Bollini and 4 other authors View PDF HTML (experimental) Abstract:Algorithmic \emph{replicability} has recently been introduced to address the need for reproducible experiments in machine learning. A \emph{replicable online learning} algorithm is one that takes the same sequence of decisions across different executions in the same environment, with high probability. We initiate the study of algorithmic replicability in \emph{constrained} MAB problems, where a learner interacts with an unknown stochastic environment for $T$ rounds, seeking not only to maximize reward but also to satisfy multiple constraints. Our main result is that replicability can be achieved in constrained MABs. Specifically, we design replicable algorithms whose regret and constraint violation match those of non-replicable ones in terms of $T$. As a key step toward these guarantees, we develop the first replicable UCB-like algorithm for \emph{unconstrained} MABs, showing that algorithms that employ the optimism in-the-face-of-uncertainty principle can be replicable, a result that we believe is of independent interest. Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2602.14580 [...