covmats.CovKernelAsLinop.solve#

CovKernelAsLinop.solve(b: ndarray[tuple[Any, ...], dtype[float64]], rtol: float = 1e-12, maxiter: int = 1000) → ndarray[tuple[Any, ...], dtype[float64]][source]#

Solve Ax = b, with A, the current covariance matrix instance.

Notes

It relies on GMRES and matrix-vector multiplications (column by column).

Parameters:

  • b (NDArrayFloat) – Array with shape (n,), (n, 1) or (n, ne), ne being the number of vectors (columns). x will have the same shape as b.

  • rtol (float, optional) – Relative tolerance for the convergence of GMRES, by default 1e-12

  • maxiter (int, optional) – Maximum number of GMRES iterations, by default 1000

Returns:

The x array with same shape as b.

Return type:

NDArrayFloat