Chebyshev HOPGD with sparse grid sampling for parameterized linear systems

We consider approximating solutions to parameterized linear systems of the form $A(\mu1,\mu2) x(\mu1,\mu2) = b$, where $(\mu1, \mu2) \in \mathbb{R}^2$. Here the matrix $A(\mu1,\mu2) \in \mathbb{R}^{n \times n}$ is nonsingular, large, and sparse and depends nonlinearly on the parameters $\mu1$ and $\mu2$. Specifically, the system arises from a discretization of a partial differential equation and $x(\mu1,\mu2) \in \mathbb{R}^n$, $b \in \mathbb{R}^n$. This work combines companion linearization with the Krylov subspace method preconditioned bi-conjugate gradient (BiCG) and a decomposition of a tensor matrix of precomputed solutions, called snapshots. As a result, a reduced order model of $x(\mu1,\mu2)$ is constructed, and this model can be evaluated in a cheap way for many values of the parameters. Tensor decompositions performed on a set of snapshots can fail to reach a certain level of accuracy, and it is not known a priori if a decomposition will be successful. Moreover, the selection of snapshots can affect both the quality of the produced model and the computation time required for its construction. This new method offers a way to generate a new set of solutions on the same parameter space at little additional cost. An interpolation of the model is used to produce approximations on the entire parameter space, and this method can be used to solve a parameter estimation problem. Numerical examples of a parameterized Helmholtz equation show the competitiveness of our approach. The simulations are reproducible, and the software is available online.