arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2310.17852cs.LGcs.AI
收藏

用于贝叶斯神经网络的函数空间贝叶斯伪核心集

Function Space Bayesian Pseudocoreset for Bayesian Neural Networks

KAIST AI(韩国科学技术院人工智能学院);AITRICS

Balhae Kim, Hyungi Lee, Juho Lee

AI总结:

本文提出在函数空间而非权重空间中构建贝叶斯伪核心集,通过匹配函数空间上的变分后验来绕过权重空间中的可扩展性和多模态问题,实验表明该方法能提升不确定性量化和鲁棒性。

摘要:

贝叶斯伪核心集是一种紧凑的合成数据集,它汇总了大规模数据集的基本信息,因此可以用作可扩展贝叶斯推断的代理数据集。通常,贝叶斯伪核心集是通过最小化以伪核心集为条件的后验与以完整数据集为条件的后验之间的散度度量来构建的。然而,评估该散度可能具有挑战性,特别是对于像深度神经网络这样具有高维参数的模型。在本文中,我们提出了一种在函数空间上运行的新型贝叶斯伪核心集构建方法。与以往在模型参数(权重)空间中构建并匹配核心集后验和完整数据后验的方法不同,我们的方法在函数空间上构建核心集后验的变分近似,并在函数空间中将其与完整数据后验进行匹配。通过直接在函数空间上工作,我们的方法可以绕过在权重空间上工作时可能出现的若干挑战,包括有限的可扩展性和多模态问题。通过各种实验,我们证明了由我们的方法构建的贝叶斯伪核心集在各种模型架构上具有增强的不确定性量化能力和更好的鲁棒性。

英文摘要:

A Bayesian pseudocoreset is a compact synthetic dataset summarizing essential information of a large-scale dataset and thus can be used as a proxy dataset for scalable Bayesian inference. Typically, a Bayesian pseudocoreset is constructed by minimizing a divergence measure between the posterior conditioning on the pseudocoreset and the posterior conditioning on the full dataset. However, evaluating the divergence can be challenging, particularly for the models like deep neural networks having high-dimensional parameters. In this paper, we propose a novel Bayesian pseudocoreset construction method that operates on a function space. Unlike previous methods, which construct and match the coreset and full data posteriors in the space of model parameters (weights), our method constructs variational approximations to the coreset posterior on a function space and matches it to the full data posterior in the function space. By working directly on the function space, our method could bypass several challenges that may arise when working on a weight space, including limited scalability and multi-modality issue. Through various experiments, we demonstrate that the Bayesian pseudocoresets constructed from our method enjoys enhanced uncertainty quantification and better robustness across various model architectures.

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。

↑