Searcharxiv⌕ Search

arXiv subjects

Xiaopei Jiao

Publications and source records attributed to Xiaopei Jiao.

2 recordsLinked to original sources

Neural Operator Approximation for Yau--Yau Nonlinear Filtering: Stability, Approximation, and Fast Computation

Since the introduction of the Kalman filter, nonlinear state estimation has become an important topic in areas such as aerospace engineering, sensor design, and target tracking. Within the optimal filtering framework described by the Duncan-Mortensen-Zakai (DMZ) equation, Yau-Yau filtering (YYF) employs an offline-online decomposition: the Fokker-Planck-Kolmogorov (FKE) equations, initialized by a prescribed set of basis functions, are solved offline, and the prior density is projected onto these basis functions during the online stage. However, this approach still faces two computational challenges: solving multiple FKEs offline can be computationally expensive, and the online projection becomes costly when filtering is performed repeatedly or in high-dimensional settings. To address these limitations, we develop a neural operator filter (NOF) that directly learns the mapping from the prior density to the posterior density, thereby eliminating both the expensive offline FKE solves and the online projection. The underlying operator is approximated using a Fourier neural operator (FNO). Numerical experiments on several challenging filtering systems verify both the effectiveness and accuracy of our method, and we further provide a rigorous theoretical analysis of its stability and approximation error.

math.OC↗

Leveraging Lie Group Symmetries to Enhance Physics-Informed Neural Networks for the Fundamental Solution of Linear PDEs

Since the introduction of deep learning for solving partial differential equations (PDEs), there has been growing interest in real-time system responses, where the kernel function plays a key role. Physics-informed neural networks (PINNs), a popular mesh-free, semi-supervised learning tool, offer high flexibility. This paper explores the integration of Lie symmetry groups with deep learning techniques to enhance the numerical solutions of fundamental PDEs. We propose a novel approach that combines PINNs and Lie group theory to address computational inefficiencies in traditional methods. By incorporating the linearized symmetric condition (LSC) derived from Lie symmetries into PINNs, we introduce a new residual loss function that requires fewer derivatives for calculation. This integration reduces computational costs and improves solution accuracy. Numerical simulations demonstrate a significant reduction in training time while maintaining accuracy. Additionally, we provide a framework for identifying invariant infinitesimal generators for arbitrary Cauchy problems. This unsupervised algorithm does not require prior numerical solutions, making it practical and efficient for various applications.

math.NA↗