arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2302.14015stat.MLcs.AIcs.LGstat.CO
收藏

CO-BED:基于贝叶斯实验设计的信息论上下文优化

CO-BED: Information-Theoretic Contextual Optimization via Bayesian Experimental Design

University of Oxford(牛津大学);Microsoft Research(微软研究院)

Desi R. Ivanova, Joel Jennings, Tom Rainforth, Cheng Zhang, Adam Foster

AI总结:

本研究从贝叶斯实验设计视角形式化上下文优化问题,提出通用模型无关框架CO-BED,用黑盒变分法结合连续松弛方案求解,在多实验中表现出媲美定制方案的竞争力。

摘要:

我们从贝叶斯实验设计的视角将上下文优化问题形式化,并提出CO-BED——一种通用的、模型无关的框架,用于利用信息论原理设计上下文实验。在构建合适的基于信息的目标函数后,我们采用黑盒变分方法,在单个随机梯度方案中同时估计该目标并优化实验设计。此外,为了在我们的框架中适配离散动作,我们提出利用连续松弛方案,该方案可自然地集成到我们的变分目标中。因此,CO-BED为广泛的上下文优化问题提供了通用的自动化解决方案。我们在多项实验中验证了其有效性,实验中CO-BED即使与定制的、特定于模型的替代方案相比,也展现出具有竞争力的性能。

英文摘要:

We formalize the problem of contextual optimization through the lens of Bayesian experimental design and propose CO-BED -- a general, model-agnostic framework for designing contextual experiments using information-theoretic principles. After formulating a suitable information-based objective, we employ black-box variational methods to simultaneously estimate it and optimize the designs in a single stochastic gradient scheme. In addition, to accommodate discrete actions within our framework, we propose leveraging continuous relaxation schemes, which can naturally be integrated into our variational objective. As a result, CO-BED provides a general and automated solution to a wide range of contextual optimization problems. We illustrate its effectiveness in a number of experiments, where CO-BED demonstrates competitive performance even when compared to bespoke, model-specific alternatives.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑