Scalable Exploration via Ensemble++

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Li, Yingru, Xu, Jiawei, Wang, Baoxiang, Luo, Zhi-Quan
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914116721967104
author Li, Yingru
Xu, Jiawei
Wang, Baoxiang
Luo, Zhi-Quan
author_facet Li, Yingru
Xu, Jiawei
Wang, Baoxiang
Luo, Zhi-Quan
contents Thompson Sampling is a principled method for balancing exploration and exploitation, but its real-world adoption faces computational challenges in large-scale or non-conjugate settings. While ensemble-based approaches offer partial remedies, they typically require prohibitively large ensemble sizes. We propose Ensemble++, a scalable exploration framework using a novel shared-factor ensemble architecture with random linear combinations. For linear bandits, we provide theoretical guarantees showing that Ensemble++ achieves regret comparable to exact Thompson Sampling with only $Θ(d \log T)$ ensemble sizes--significantly outperforming prior methods. Crucially, this efficiency holds across both compact and finite action sets with either time-invariant or time-varying contexts without configuration changes. We extend this theoretical foundation to nonlinear rewards by replacing fixed features with learnable neural representations while preserving the same incremental update principle, effectively bridging theory and practice for real-world tasks. Comprehensive experiments across linear, quadratic, neural, and GPT-based contextual bandits validate our theoretical findings and demonstrate Ensemble++'s superior regret-computation tradeoff versus state-of-the-art methods.
format Preprint
id arxiv_https___arxiv_org_abs_2407_13195
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scalable Exploration via Ensemble++
Li, Yingru
Xu, Jiawei
Wang, Baoxiang
Luo, Zhi-Quan
Machine Learning
Artificial Intelligence
Human-Computer Interaction
Information Theory
Thompson Sampling is a principled method for balancing exploration and exploitation, but its real-world adoption faces computational challenges in large-scale or non-conjugate settings. While ensemble-based approaches offer partial remedies, they typically require prohibitively large ensemble sizes. We propose Ensemble++, a scalable exploration framework using a novel shared-factor ensemble architecture with random linear combinations. For linear bandits, we provide theoretical guarantees showing that Ensemble++ achieves regret comparable to exact Thompson Sampling with only $Θ(d \log T)$ ensemble sizes--significantly outperforming prior methods. Crucially, this efficiency holds across both compact and finite action sets with either time-invariant or time-varying contexts without configuration changes. We extend this theoretical foundation to nonlinear rewards by replacing fixed features with learnable neural representations while preserving the same incremental update principle, effectively bridging theory and practice for real-world tasks. Comprehensive experiments across linear, quadratic, neural, and GPT-based contextual bandits validate our theoretical findings and demonstrate Ensemble++'s superior regret-computation tradeoff versus state-of-the-art methods.
title Scalable Exploration via Ensemble++
topic Machine Learning
Artificial Intelligence
Human-Computer Interaction
Information Theory
url https://arxiv.org/abs/2407.13195