Applications of Mathematics最新文献

This work introduces an adaptive mesh refinement technique for hierarchical hybrid grids with the goal to reach scalability and maintain excellent performance on massively parallel computer systems. On the block-structured hierarchical hybrid grids, this is accomplished by using classical, unstructured refinement only on the coarsest level of the hierarchy, while keeping the number of structured refinement levels constant over the whole domain. This leads to a compromise, where the excellent performance characteristics of hierarchical hybrid grids can be maintained at the price that the flexibility of generating locally refined meshes is constrained. Furthermore, the mesh adaptivity often relies on a posteriori error estimators or error indicators, which tend to become computationally expensive. Again, with the goal of preserving scalability and performance, a method is proposed that leverages the grid hierarchy and the full multigrid scheme. Utilizing the sequence of approximations on the nested hierarchy of grids permits the computation of a cheap error estimator that is well-suited for large-scale parallel computing. We present the theoretical foundations for both global and local error estimates, and present a rigorous analysis of their effectivity. The proposed method, including the error estimator and the adaptive coarse grid refinement, is implemented in the finite element framework HyTeG. Extensive numerical experiments are conducted to validate the effectiveness, as well as performance and scalability.

We investigate the Dirichlet boundary control of the Laplace equation, considering the control in H^1/2(∂Ω), which is the natural space for Dirichlet data when the state belongs to H¹(Ω) The cost of the control is measured in the H^1/2(∂Ω) norm that also plays the role of the regularization term. We discuss regularization and finite element error estimates enabling us to derive an optimal relation between the finite element mesh size h and the regularization parameter ϱ, balancing the energy cost for the control and the accuracy of the approximation of the desired state. This relationship is also crucial in designing efficient solvers. We also discuss additional box constraints imposed on the control and the state. Our theoretical findings are complemented by numerical examples, including one example with box constraints.

The numerical solution of linear systems obtained as a result of discretization of a spectral fractional diffusion problem is studied. The finite element method is applied to the considered boundary value problem. The system matrix is a fractional power of the product of the inverse of the mass matrix and the stiffness matrix. The matrix thus defined is symmetric and positive definite (SPD) with respect to the inner product associated with the mass matrix, but is dense, which is consistent with the nonlocal nature of fractional diffusion. The presented results are in the spirit of the BURA (Best Uniform Rational Approximation) method. BURA reduces numerical solution of the dense linear system to the solution of k systems with sparse SPD diffusion-reaction matrices, where k is the degree of rational approximation. We prove the existence of algebraic multilevel iteration (AMLI) methods for preconditioning such type of emergent matrices that satisfy the conditions for optimal computational complexity. Both multiplicative and additive AMLI preconditioners have been developed, determining the minimum possible degree θ of the hierarchical θ-refinement of the mesh.

The mortar method is a powerful technique to enforce constraints between non-conforming discretizations by introducing a set of Lagrange multipliers on the connecting interface. Usually, the multipliers are not obtained explicitly because they can be eliminated with the aid of the so-called mortar interpolation operator. However, their explicit computation becomes essential when the contact constraint is governed by some non-linear law, and in this situation it is necessary to guarantee that discrete spaces of the primary variables and multipliers are inf-sup stable. In this work, we investigate the issue of inf-sup stability when using various families of piecewise linear and piecewise constant multipliers. The focus is on the role of the mesh resolution and the enforcement of boundary conditions, which are important factors in practical applications. Then, we develop a stabilized formulation for piecewise-constant multipliers inspired by the framework of minimal stabilization. The effectiveness of the proposed approach is demonstrated through numerical benchmarks and examples.

We present a unified framework for estimating stochastic parameters in general variational problems. This nonlinear inverse problem is formulated as a stochastic optimization problem using the output least-squares (OLS) objective, which minimizes the discrepancy between observed data and the computed solution. A key challenge in OLS-based formulations is the efficient computation of first- and second-order derivatives of the OLS functional, which depend on the corresponding derivatives of the parameter-to-solution map often costly and difficult to evaluate, especially in stochastic settings. To address this, we develop a rigorous computational approach based on first- and second-order adjoint methods for inverse problems governed by stochastic variational problems. Specifically, we propose a new first-order adjoint method for computing the gradient of the OLS objective and introduce two novel second-order adjoint methods for Hessian evaluation. A stochastic Galerkin discretization framework is employed, enabling efficient implementation of the adjoint-based derivative computations. Numerical experiments demonstrate the accuracy and efficiency of the proposed computational framework.

The accurate computation of eigenfunctions corresponding to tightly clustered Laplacian eigenvalues remains an extremely difficult problem. Using the shape difference quotient of eigenvalues, we propose a stable computation method for the eigenfunctions of clustered eigenvalues caused by domain perturbation.

This study introduces an accelerated gradient descent method based on a non-monotone backtracking line search scheme. A simple adaptive quadratic model is enhanced by utilizing a real, positive definite scalar matrix derived from the Taylor expansion of the objective function, rather than relying on the exact Hessian. The global and superlinear convergence of the defined model is established under appropriate conditions. Numerical experiments on a set of standard unconstrained optimization problems and image restoration problems show that the new algorithm outperforms other comparable methods in terms of efficiency and robustness.

This paper addresses the identification of the leak bound function g in the Stokes system with threshold leak boundary conditions, where g varies spatially. The state problem is solved using the dual formulation of the algebraic system, and the resulting optimization problem is formulated as a nonsmooth optimization problem. We establish the existence of solutions for both the continuous and discrete formulations of the problem. The theoretical developments are complemented by numerical experiments, which compare the performance of the nonsmooth optimization approach with traditional regularization-based methods and global optimization techniques.

We propose and analyze three kinds of two-grid penalty Arrow-Hurwicz (A-H) iterative finite element methods for the stationary incompressible magnetohydrodynamic (MHD) equations, which adopt the existing A-H iterative method to obtain the coarse mesh solution, and then correct the solution by three different one-step schemes (Oseen type, Stokes type and Newton type) with the usual penalty method on the fine mesh. These methods combine the A-H iterative method, the penalty method and the two-grid strategy, maintaining the advantage of three methods and overcoming some of their limitations. Rigorous analysis of the optimal error estimate and stability for three methods are provided. Ample numerical experiments are reported to validate the theoretical results and the efficiency of the numerical schemes.

In the linear regression model, the standard coefficient of determination R² and its weighted counterpart are commonly used to assess the quality of the linear fit. However, both metrics are susceptible to the influence of outliers and heteroskedasticity within the dataset. This paper introduces a robust version of R², based on the least weighted squares (LWS) estimator, and examines its statistical properties in detail. We investigate the impact of data quantization on R² and its robust variants, and propose a hypothesis test for assessing the equality of expected values between two R² versions. Numerical experiments on 29 publicly available datasets reveal that confidence intervals for the LWS-based coefficient of determination are generally narrower than those for existing measures, especially in homoskedastic settings. In contrast, under heteroskedasticity, narrower intervals do not necessarily imply greater robustness, highlighting the nuanced behavior of these estimators. The comparison with the well-known least trimmed squares (LTS) estimator underscores the promise of the LWS approach, which exhibits favorable efficiency properties and more reliable interval estimation in many practical scenarios.