Generalized Neyman Allocation for Locally Minimax Optimal Best-Arm Identification

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autore principale: Kato, Masahiro
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929694160453632
author Kato, Masahiro
author_facet Kato, Masahiro
contents This study investigates an asymptotically locally minimax optimal algorithm for fixed-budget best-arm identification (BAI). We propose the Generalized Neyman Allocation (GNA) algorithm and demonstrate that its worst-case upper bound on the probability of misidentifying the best arm aligns with the worst-case lower bound under the small-gap regime, where the gap between the expected outcomes of the best and suboptimal arms is small. Our lower and upper bounds are tight, matching exactly including constant terms within the small-gap regime. The GNA algorithm generalizes the Neyman allocation for two-armed bandits (Neyman, 1934; Kaufmann et al., 2016) and refines existing BAI algorithms, such as those proposed by Glynn & Juneja (2004). By proposing an asymptotically minimax optimal algorithm, we address the longstanding open issue in BAI (Kaufmann, 2020) and treatment choice (Kasy & Sautmann, 202) by restricting a class of distributions to the small-gap regimes.
format Preprint
id arxiv_https___arxiv_org_abs_2405_19317
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Generalized Neyman Allocation for Locally Minimax Optimal Best-Arm Identification
Kato, Masahiro
Machine Learning
Artificial Intelligence
Econometrics
Methodology
This study investigates an asymptotically locally minimax optimal algorithm for fixed-budget best-arm identification (BAI). We propose the Generalized Neyman Allocation (GNA) algorithm and demonstrate that its worst-case upper bound on the probability of misidentifying the best arm aligns with the worst-case lower bound under the small-gap regime, where the gap between the expected outcomes of the best and suboptimal arms is small. Our lower and upper bounds are tight, matching exactly including constant terms within the small-gap regime. The GNA algorithm generalizes the Neyman allocation for two-armed bandits (Neyman, 1934; Kaufmann et al., 2016) and refines existing BAI algorithms, such as those proposed by Glynn & Juneja (2004). By proposing an asymptotically minimax optimal algorithm, we address the longstanding open issue in BAI (Kaufmann, 2020) and treatment choice (Kasy & Sautmann, 202) by restricting a class of distributions to the small-gap regimes.
title Generalized Neyman Allocation for Locally Minimax Optimal Best-Arm Identification
topic Machine Learning
Artificial Intelligence
Econometrics
Methodology
url https://arxiv.org/abs/2405.19317