Abstract:We present an efficient matrix-free point spread function (PSF) method for approximating operators that have locally supported non-negative integral kernels. The method computes impulse responses at scattered points, and interpolates these impulse responses to approximate integral kernel entries. Impulse responses are computed by applying the operator to Dirac comb batches of point sources, which are chosen via an ellipsoid packing procedure. Evaluation of kernel entries allows us to construct a hierarchical matrix approximation of the operator, which is used for further matrix computations. We illustrate the end-to-end method on a blur problem, then use the method to build preconditioners for the Hessian in two inverse problems governed by partial differential equations (PDEs): inversion for the basal friction coefficient in an ice sheet flow problem and for the initial condition in an advective-diffusive transport problem. While for many ill-posed inverse problems the Hessian of the data misfit term exhibits a low rank structure, and hence a low rank approximation is suitable, for many problems of practical interest the numerical rank of the Hessian is still large. But Hessian impulse responses typically become more local as the numerical rank increases, which benefits the PSF method. Numerical results reveal that the PSF preconditioner clusters the spectrum of the preconditioned Hessian near one, yielding roughly 5x-10x reductions in the required number of PDE solves, as compared to regularization preconditioning and no preconditioning. We also present a numerical study for the influence of various parameters (that control the shape of the impulse responses) on the effectiveness of the advection-diffusion Hessian approximation. The results show that the PSF-based preconditioners are able to form good approximations of high-rank Hessians using a small number of operator applications.

What problem does this paper attempt to address?

This paper aims to solve the problem of how to efficiently approximate high - rank Hessian matrices with non - negative integral kernels of local support. Specifically, the article proposes a method based on the Point Spread Function (PSF) to approximate high - rank Hessian matrices that appear in inverse problems constrained by partial differential equations (PDEs). The following are the main problems and solutions in this study: ### 1. Research Background and Problem Description In many practical applications, especially in inverse problems constrained by partial differential equations, the numerical rank of the Hessian matrix may be large, which makes low - rank approximation methods inapplicable. For example, in ice - sheet flow problems and convection - diffusion transport problems, the numerical rank of the Hessian matrix is high, resulting in a large number of PDE solutions required by low - rank approximation methods and high computational costs. ### 2. Proposed Method To solve the above problems, the author proposes a PSF - based method that can approximate high - rank Hessian matrices with a small number of operations. The specific steps are as follows: - **Impulse Response Calculation**: Calculate the impulse responses of these point sources by applying the operator to the Dirac comb function (related to a batch of point sources). - **Ellipsoid Packing Algorithm**: Select a batch of non - overlapping support ellipsoids to maximize the number of impulse responses per batch. - **Interpolation and Approximation**: Use radial basis function interpolation to approximate the transformed and scaled impulse responses as entries of the operator integral kernel. - **Hierarchical Matrix (H - matrix) Construction**: Utilize the fast H - matrix method for further matrix calculations, such as matrix - vector multiplication, matrix decomposition, etc. ### 3. Advantages of the Method Compared with traditional low - rank approximation methods, the PSF method has the following advantages: - **Higher Precision**: It can obtain a more accurate Hessian approximation with fewer operations. - **Better Locality**: It takes advantage of the local characteristics of the Hessian impulse response and improves the effectiveness of the approximation. - **Reduced Computational Cost**: Compared with the randomized H - matrix construction method, the number of required operations is significantly reduced. ### 4. Experimental Results The author verifies the effectiveness of the PSF method through numerical experiments. The experimental results show that the PSF method can significantly reduce the number of required PDE solutions while maintaining high approximation accuracy. In addition, the influence of different parameters (such as impulse response shape and Hessian rank) on the approximation effect is also studied, indicating that the PSF method can form a good high - rank Hessian approximation with a relatively small number of operations. ### Summary This paper mainly solves the problem of how to efficiently approximate high - rank Hessian matrices in PDE - constrained inverse problems. By introducing the PSF - based method, the author provides a new and efficient approximation means that can significantly reduce the computational cost while ensuring accuracy. This method has important application value in dealing with complex inverse problems.

Point spread function approximation of high rank Hessians with locally supported non-negative integral kernels

A new interpolated pseudodifferential preconditioner for the Helmholtz equation in heterogeneous media

Non-equispaced Fourier Neural Solvers for PDEs

On Edge Multiscale Space based Hybrid Schwarz Preconditioner for Helmholtz Problems with Large Wavenumbers

A RELAXED HSS PRECONDITIONER FOR SADDLE POINT PROBLEMS FROM MESHFREE DISCRETIZATION

A Scalable Interior-Point Gauss-Newton Method for PDE-Constrained Optimization with Bound Constraints

Pseudospectral method for solving PDEs using Matrix Product States

Preconditioned low-rank Riemannian optimization for linear systems with tensor product structure

Fast and Scalable FFT-Based GPU-Accelerated Algorithms for Hessian Actions Arising in Linear Inverse Problems Governed by Autonomous Dynamical Systems

Noniterative Least-Squares Reverse Time Migration Based on an Efficient Asymptotic Hessian/point-spread Function Estimation

Two-level Restricted Additive Schwarz preconditioner based on Multiscale Spectral Generalized FEM for Heterogeneous Helmholtz Problems

Optimal operator preconditioning for pseudodifferential boundary problems

Application of shifted-Laplace preconditioners for heterogenous Helmholtz equation- part 1: Data modelling

A matrix algebra approach to approximate Hessians

Efficient additive Schwarz preconditioning for hypersingular integral equations on locally refined triangulations

A novel direct Helmholtz solver in inhomogeneous media based on the operator Fourier transform functional calculus

A -Preconditioner for Space Fractional Diffusion Equation with Non-separable Variable Coefficients

Hierarchical Super-Localized Orthogonal Decomposition Method

Two-level hybrid Schwarz Preconditioners for The Helmholtz Equation with high wave number

A Relaxed Deteriorated Pss Preconditioner for Nonsymmetric Saddle Point Problems from the Steady Navier-Stokes Equation

Sparse Gaussian processes for solving nonlinear PDEs