Source code for pycif.plugins.minimizers.congrad.minimize

import copy

import numpy as np


from logging import info




def minimize(self, finit, gradinit, chi0, **kwargs):
    """Run the CONGRAD Lanczos minimisation and return the optimal iterate.

    Args:
        self (Plugin): CONGRAD minimizer plugin instance.
        finit (float): initial cost function value :math:`J(\chi_0)`.
        gradinit (np.ndarray): initial gradient :math:`\\nabla J(\chi_0)`,
            shape ``(n,)``.
        chi0 (np.ndarray): initial iterate :math:`\chi_0`, shape ``(n,)``.
        **kwargs: forwarded to the simulator.

    Returns:
        np.ndarray: optimal iterate :math:`\chi_\mathrm{opt}`, shape ``(n,)``.
    """

    # Initializing options (and filling missing values with default)
    self = self.check_options(chi0, **kwargs)

    # Running CONGRAD
    lanczvect0 = copy.deepcopy(gradinit)
    xopt, gradopt, preduc, pevecs, iiter = self.congrad(
        chi0, gradinit, lanczvect0, **kwargs
    )

    # Save pevecs as uncertainty reductions
    if self.save_uncertainties:
        self.pevecs = pevecs

    # Final verbose and output
    towrite = """
        CONGRAD:
            number of iterations: {}
            achieved relative reduction of the gradient: {}
        """.format(
        iiter, preduc
    )

    info(towrite)

    r1 = np.sqrt(np.dot(xopt, xopt))
    r2 = np.sqrt(np.dot(gradopt, gradopt))

    info("norm of x = " + str(r1))
    info("norm of g = " + str(r2))

    return xopt