An Anytime Algorithm for Good Arm Identification
In good arm identification (GAI), the goal is to identify one arm whose average performance exceeds a given threshold, referred to as a good arm, if it exists. Few works have studied GAI in the fixed-budget setting when the sampling budget is fixed beforehand, or in the anytime setting, when a recommendation can be asked at any time. We propose APGAI, an anytime and parameter-free sampling rule for GAI in stochastic bandits. APGAI can be straightforwardly used in fixed-confidence and fixed-budget settings. First, we derive upper bounds on its probability of error at any time. They show that adaptive strategies can be more efficient in detecting the absence of good arms than uniform sampling in several diverse instances. Second, when APGAI is combined with a stopping rule, we prove upper bo
Marc Jourdan, Andrée Delahaye-Duriez, Clémence Réda; 27(19):1−90, 2026.
Abstract
In good arm identification (GAI), the goal is to identify one arm whose average performance exceeds a given threshold, referred to as a good arm, if it exists. Few works have studied GAI in the fixed-budget setting when the sampling budget is fixed beforehand, or in the anytime setting, when a recommendation can be asked at any time. We propose APGAI, an anytime and parameter-free sampling rule for GAI in stochastic bandits. APGAI can be straightforwardly used in fixed-confidence and fixed-budget settings. First, we derive upper bounds on its probability of error at any time. They show that adaptive strategies can be more efficient in detecting the absence of good arms than uniform sampling in several diverse instances. Second, when APGAI is combined with a stopping rule, we prove upper bounds on the expected sampling complexity, holding at any confidence level. Finally, we show the good empirical performance of APGAI on synthetic and real-world data. Our work offers an extensive overview of the GAI problem in all settings.
[abs][pdf][bib]
Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Laws & Regulation
Sources: threat actors stole Cisco source code by breaching its internal development environment using credentials from a recent Trivy supply chain attack (Lawrence Abrams/BleepingComputer)
Lawrence Abrams / BleepingComputer : Sources: threat actors stole Cisco source code by breaching its internal development environment using credentials from a recent Trivy supply chain attack — Cisco has suffered a cyberattack after threat actors used stolen credentials from the recent Trivy supply chain attack to breach …
NYC Mayor Zohran Mamdani allows city agencies to use TikTok, with some restrictions, reversing a 2023 ban his predecessor enacted over data security concerns (Chris Sommerfeldt/Politico)
Chris Sommerfeldt / Politico : NYC Mayor Zohran Mamdani allows city agencies to use TikTok, with some restrictions, reversing a 2023 ban his predecessor enacted over data security concerns — The @nycmayor handle had been inactive since 2023 when former Mayor Eric Adams banned TikTok from being used on city government devices due …
Mark Zuckerberg offered to 'help' Elon Musk with DOGE in 2025
Elon Musk and Mark Zuckerberg have a complicated history. In 2023, the two vowed to fight each other in a cage match that never happened . But by early 2025, when both were cozying up to the newly-elected President Donald Trump, they were apparently on more friendly terms. In February of that year, Zuckerberg texted Musk approvingly about his work with the now-defunct Department of Government Efficiency (DOGE). "Looks like DOGE is making progress," the Meta CEO texted. "I've got our teams on alert to take down content doxxing or threatening the people on your team. Let me know if there's anything else I can do to help." The texts, which were published Friday in court documents as part of Musk's lawsuit against Sam Altman and OpenAI, are dated February 3, 2025. That's just a few weeks after
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!