Bayesian Inverse Problems for EUV reflectometry with regularized JKO scheme
基于正则化JKO格式的EUV反射测量贝叶斯反问题
Abstract
We study a Bayesian inverse problem in EUV reflectometry, where optical constants and layer thicknesses are inferred from angle- and wavelength-resolved reflection measurements. The physical process is modeled and evaluated using an efficient forward model based on the transfer matrix method. As an alternative to computationally expensive sampling-based inference, we employ an Eulerian, entropically regularized Jordan-Kinderlehrer-Otto (JKO) scheme that minimizes the Kullback--Leibler divergence via proximal steps in the Wasserstein space. Using the Benamou--Brenier dynamic formulation and a Hopf-Cole transform, each JKO step is reformulated as a pair of heat equations coupled only through their initial and terminal conditions, enabling an efficient fixed-point solver. The resulting high-dimensional densities are discretized on tensor-product grids and represented in tensor-train format, exploiting Kronecker structure for scalable computation. Preconditioning based on estimated posterior moments further improves tractability. We test the method with numerical experiments on EUV reflectometry data for a ruthenium--silicon ($\ce{Ru}$ on $\ce{Si}$) specimen and compare the learned tensor-train posterior to ensemble Markov chain Monte Carlo (MCMC) reference samples. We demonstrate a use case in which we are able to reuse the tensor-train posterior to obtain samples for shifted target distributions with less computational effort than a comparable MCMC method.
摘要
我们研究EUV反射测量中的贝叶斯反问题，其中从角度和波长分辨的反射测量中推断光学常数和层厚度。该物理过程使用基于传输矩阵法的高效正演模型进行建模和评估。作为计算昂贵的基于采样的推断的替代方案，我们采用欧拉型、熵正则化的Jordan-Kinderlehrer-Otto（JKO）格式，该格式通过Wasserstein空间中的近端步骤最小化Kullback--Leibler散度。利用Benamou--Brenier动态表述和Hopf-Cole变换，每个JKO步被重新表述为仅通过初始和终端条件耦合的一对热方程，从而实现高效的定点求解器。所得高维密度在张量积网格上离散，并以张量列格式表示，利用Kronecker结构实现可扩展计算。基于估计后验矩的预条件进一步改善了可处理性。我们在钌--硅（$\ce{Ru}$ on $\ce{Si}$）样品的EUV反射测量数据上进行数值实验，以测试该方法，并将学习到的张量列后验与集合马尔可夫链蒙特卡洛（MCMC）参考样本进行比较。我们展示了一个用例：能够重用张量列后验，以比可比MCMC方法更少的计算量获得偏移目标分布的样本。
Why it matters
This paper sits squarely in the tensor-train side of the researcher's interests: it solves a high-dimensional Bayesian inverse problem by evolving probability densities in the Wasserstein space (JKO scheme), where each proximal step reduces to a pair of coupled heat equations, and it discretizes the resulting high-dimensional densities on tensor-product grids represented in tensor-train format, explicitly exploiting Kronecker structure for scalability. The combination of TT compression, structure-exploiting fixed-point solvers, and moment-based preconditioning for high-dimensional PDE-driven density evolution is directly transferable to kinetic/Fokker-Planck-type problems and to low-rank solvers for high-dimensional evolution equations. The paper also provides a concrete, non-trivial validation (EUV reflectometry with MCMC reference comparison, plus reuse of the TT posterior for shifted targets), which is valuable evidence for the practical competitiveness of TT-based approaches.
推荐理由
该文正好落在研究者关注的张量列（TT）方向：它以 Wasserstein 空间中的 JKO 格式求解高维贝叶斯反问题，每个近端步化为耦合的热方程对，并将所得高维密度在张量积网格上以张量列格式离散，显式利用 Kronecker 结构实现可扩展计算。TT 压缩、结构利用的不动点求解器与基于矩的预条件相结合，可直接迁移到动理学/Fokker-Planck 型问题以及高维发展方程的低秩求解。文章还给出了较为实质的验证（EUV 反射测量，与系综 MCMC 对比，并复用 TT 后验处理平移目标分布），为 TT 方法的实际竞争力提供了有价值的证据。
Key contribution
Proposes an Eulerian, entropically regularized JKO scheme for Bayesian inverse problems, reformulating each proximal step via the Benamou-Brenier formulation and Hopf-Cole transform into two coupled heat equations solved by a fixed-point iteration. High-dimensional densities are discretized on tensor-product grids and stored in tensor-train format with Kronecker-structure exploitation and moment-based preconditioning; validated on EUV reflectometry data with comparison against ensemble MCMC and reuse of the TT posterior for shifted target distributions.
核心贡献
提出用于贝叶斯反问题的欧拉型、熵正则化 JKO 格式，借助 Benamou-Brenier 表述与 Hopf-Cole 变换，将每个近端步重写为两个仅通过初末条件耦合的热方程，并用不动点迭代求解。高维密度在张量积网格上离散并以张量列格式存储，利用 Kronecker 结构并采用基于后验矩的预条件；在 EUV 反射测量数据上验证，与系综 MCMC 参考样本比较，并展示了复用 TT 后验以更低代价获得平移目标分布样本的用例。