Deep Adaptive Dimension Reduction for Bayesian Inference in Inverse Problems

2026-05-28

AI Summary

This work addresses the challenges of posterior complexity, expensive forward models, and prior misspecification in high-dimensional inverse problems governed by partial differential equations. To this end, the authors propose a deep adaptive dimensionality reduction framework for Bayesian inference that integrates nonlinear dimensionality reduction with dual normalizing flows. This approach overcomes the bijectivity constraint inherent in conventional normalizing flows, yielding a tighter evidence lower bound, and incorporates an iterative prior updating mechanism that requires no manual hyperparameter tuning. A surrogate model based on adaptively tuned Fourier neural operators is constructed to enable closed-loop co-optimization between posterior inference and surrogate modeling. Experiments on a 100-dimensional Rosenbrock benchmark and three PDE-based inverse problems demonstrate that the proposed method achieves inference accuracy comparable to or better than state-of-the-art approaches—including MCMC, UKI, and SVGD—particularly in high-dimensional and high-noise regimes.

Abstract

Solving high-dimensional PDE-governed inverse problems is often challenging due to complex non-Gaussian posterior distributions, expensive forward model evaluations, and misspecified prior information. To address these issues, we propose a deep adaptive dimension-reduction Bayesian inference framework based on the Variational Flow (VF) model. Since standard normalizing flows are restricted by bijective mappings and cannot directly reduce dimensions, VF overcomes this limitation by integrating VAE-based nonlinear dimension reduction with dual normalizing flows for the latent prior and encoder. This design provides a strictly higher evidence lower bound than VAE and allows more flexible approximation of complex posterior distributions. We further introduce an iterative prior updating strategy that gradually moves the prior mean toward high-probability posterior regions, avoiding manual prior tuning. These components form a closed adaptive loop together with an adaptively fine-tuned Fourier Neural Operator (FNO) surrogate: VF generates posterior-concentrated samples to refine the surrogate, while the updated surrogate further improves posterior inference. Numerical experiments on a 100-dimensional Rosenbrock problem and three standard PDE-governed inverse problems show that our method delivers competitive or superior accuracy compared with MCMC, UKI, and SVGD baselines across all tested configurations, with the most pronounced advantages emerging in challenging scenarios such as high-noise observations and high-dimensional parameter spaces.

Problem

• inverse problems
• Bayesian inference
• high-dimensional PDE
• non-Gaussian posterior
• prior misspecification

Innovation

• Variational Flow
• Dimension Reduction
• Bayesian Inference
• Fourier Neural Operator
• Inverse Problems