arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2205.06342stat.MLcs.LGstat.COstat.ME
收藏

函数空间中的广义变分推断:高斯测度满足贝叶斯深度学习

Generalized Variational Inference in Function Spaces: Gaussian Measures meet Bayesian Deep Learning

University of Oxford(牛津大学);Amazon(亚马逊)

Veit D. Wild, Robert Hu, Dino Sejdinovic

摘要:

我们开发了一个框架,广义变分推断在无限维函数空间,并使用它来构建一个方法称为高斯Wasserstein推理(GWI)。GWI利用平方可积函数的希尔伯特空间上的高斯测度之间的瓦瑟斯坦距离,以便使用易于处理的优化标准来确定变分后验,并避免标准变分函数空间推断中出现的病态。GWI的一个令人兴奋的应用是能够在GWI的变分参数化中使用深度神经网络,将其卓越的预测性能与类似于高斯过程的原则性不确定性量化相结合。所提出的方法在几个基准数据集上获得了最先进的性能。

英文摘要:

We develop a framework for generalized variational inference in infinite-dimensional function spaces and use it to construct a method termed Gaussian Wasserstein inference (GWI). GWI leverages the Wasserstein distance between Gaussian measures on the Hilbert space of square-integrable functions in order to determine a variational posterior using a tractable optimisation criterion and avoids pathologies arising in standard variational function space inference. An exciting application of GWI is the ability to use deep neural networks in the variational parametrisation of GWI, combining their superior predictive performance with the principled uncertainty quantification analogous to that of Gaussian processes. The proposed method obtains state-of-the-art performance on several benchmark datasets.

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。

↑