power_method#

mrinufft.operators.base.power_method(max_iter: int, operator: FourierOperatorBase | Callable, norm_func: Callable | None = None, x: ndarray[tuple[int, ...], dtype[_ScalarType_co]] | None = None) → tuple[floating | ndarray[tuple[int, ...], dtype[_ScalarType_co]], ndarray[tuple[int, ...], dtype[_ScalarType_co]]][source]#

Power method to find the Lipschitz constant of an operator.

Parameters:

  • max_iter (int) – Maximum number of iterations

  • operator (FourierOperatorBase or child class or Callable) – NUFFT Operator of which to estimate the lipchitz constant. If it is Callable, it should implement the AHA operation.

  • norm_func (callable, optional) – Function to compute the norm , by default np.linalg.norm. Change this if you want custom norm, or for computing on GPU.

  • x (array_like, optional) – Initial value to use, by default a random numpy array is used.

Returns:

  • x_new_norm (float or NDArray) – The maximum eigen value

  • x_new (NDArray) – The eigen vector associated with maximum eigen value

Return type:

tuple[floating | ndarray[tuple[int, …], dtype[_ScalarType_co]], ndarray[tuple[int, …], dtype[_ScalarType_co]]]