Abstract:We consider a randomised implementation of the finite element method (FEM) for elliptic partial differential equations on high-dimensional models. This is motivated by applications where model predictions are essential for real-time process diagnostics. In these circumstances it is imperative to expedite prediction without a significant compromise in the model's fidelity, which in turn relies on the rapid assembly and solution of the associated system of equations typically at the many-query context. Our approach involves converting the solution of the linear, symmetric positive definite FEM system into an over-determined least squares problem, whose solution is then projected onto a low-dimensional subspace. The resulting low-dimensional system can be effectively sketched as a product of two high-dimensional matrices using a parameter-dependent non-uniform sampling distribution, utilising only a small subset of the model's parameters. Although different to the optimal sampling distributions based on the statistical leverage-scores of the rows of the matrices, we show that the distance between them shrinks for an appropriate choice of the projection subspace. For the approximate solution we bound the incurring errors due to the projection, subspace approximation and sketching and show that the overall error is dominated by the condition number of the projected stiffness matrix. Our approach is tested on simulations on the Dirichlet and Neumann problems for the steady-state diffusion equation. The results show that our approach has on average a tenfold improvement on the computational times compared to the classical deterministic framework at the expense of a moderately small approximation error.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is: **In high - dimensional models, how to accelerate the finite element method (FEM) solution of elliptic partial differential equations while maintaining the fidelity of the model**. Specifically, the author proposes a randomized finite element method based on row sampling to achieve fast, real - time prediction, especially in a multi - query environment.
### Problem Background
In many application scenarios, such as online calibration of sensor networks or control of manufacturing processes, fast, real - time simulation is required. However, using the traditional deterministic framework for large three - dimensional model simulation is a huge challenge and usually requires relying on high - performance computing infrastructure. Therefore, reducing the computational time for forward problem solving has always been a key issue in model order reduction and statistical inference algorithms.
### Main Contributions of the Paper
1. **Randomization Method**: The author transforms the linear symmetric positive - definite FEM system into an over - determined least - squares problem and projects it onto a low - dimensional subspace. By using a parameter - dependent non - uniform sampling distribution, a small subset of model parameters is used to effectively sketch the product of high - dimensional matrices.
2. **Error Analysis**: For the approximate solution, the author bounds the errors caused by projection, subspace approximation, and sketching, and proves that the overall error is mainly dominated by the condition number of the projected stiffness matrix.
3. **Numerical Experiments**: This method was tested in the simulation of the steady - state diffusion equation for Dirichlet and Neumann problems. The results show that the average computation time is ten times faster than that of the classical deterministic framework, while the approximation error is small.
### Key Formulas
- Elliptic Partial Differential Equation:
\[
-\nabla\cdot(p(x)\nabla u(x)) = f(x)\quad\text{in }\Omega\subset\mathbb{R}^3
\]
where the boundary conditions are:
\[
u = g(D)\quad\text{on }\partial\Omega_D,\quad\nabla u\cdot\hat{n}=g(N)\quad\text{on }\partial\Omega_N
\]
- Least - Squares Problem:
\[
\hat{u}_{\text{ls}}=\arg\min_{u\in\mathbb{R}^n}\|Yu - Z^{-\frac{1}{2}}(D^T)^{\dagger}b\|^2
\]
- Projected Least - Squares Problem:
\[
r=\arg\min_{r\in\mathbb{R}^{\rho}}\|Xr - Z^{-\frac{1}{2}}(D^T)^{\dagger}b\|^2
\]
- Construction of the Randomized Matrix \(\hat{G}\):
\[
\hat{G}=X^TSS^TX=\frac{1}{c}\sum_{t = 1}^c\frac{1}{\xi_{r_t}}X^T_{(r_t)*}X_{(r_t)*}
\]
Through these methods, the author has achieved a significant acceleration of the finite element solution of elliptic partial differential equations while maintaining the fidelity of the model.