We consider the problem of uncertainty quantification (UQ) for an unknown low-rank matrix $\mathbf{X}$, given a partial and noisy observation of its entries. This quantification of uncertainty is essential for many real-world problems, including image processing, satellite imaging, and seismology, providing a principled framework for validating scientific conclusions and guiding decision-making. However, existing literature has largely focused on the completion (i.e., point estimation) of the matrix $\mathbf{X}$, with little work on exploring the uncertainty of such estimates. To this end, we propose in this work a new Bayesian modeling framework, called BayeSMG, which parametrizes the unknown $\mathbf{X}$ via its underlying row and column subspaces. This Bayesian subspace parametrization allows for efficient posterior inference on matrix subspaces (which represent interpretable phenomena in many applications), which can then be leveraged for improved matrix recovery. We demonstrate the effectiveness of BayeSMG over existing Bayesian matrix recovery methods in extensive numerical experiments and a seismic sensor network application.
翻译:我们认为,鉴于对一个未知的低级矩阵($\mathbf{X})进行局部和紧张的观察,不确定性的量化问题(UQ)是一个未知的低级矩阵($\mathbf{X})的问题。这种不确定性的量化对于许多现实世界问题至关重要,包括图像处理、卫星成像和地震学,为验证科学结论和指导决策提供了一个原则框架。然而,现有文献主要侧重于矩阵($\mathbf{X})的完成(即点估测),很少探讨这种估算的不确定性。为此,我们提议在这项工作中建立一个称为Bayesian模型的新框架,称为BayeSMG, 它通过其基本行和柱体次空间对未知的美元进行假称。这种巴耶斯次空间的相近光谱化使得对矩阵子空间(在许多应用中代表可解释的现象)的有效外推力得以利用,从而改进矩阵的恢复。我们通过广泛的数字实验和地震网络应用,展示了BaySMG相对于现有的贝斯矩阵恢复方法的有效性。