yango.py 3.65 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
#
# Copyright(C) 2013-2018 Max-Planck-Society
#
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik
# and financially supported by the Studienstiftung des deutschen Volkes.

from __future__ import division
from .minimizer import Minimizer
Martin Reinecke's avatar
Martin Reinecke committed
21
from ..logger import logger
22
from .line_search_strong_wolfe import LineSearchStrongWolfe
23
import numpy as np
24 25 26 27


class Yango(Minimizer):
    """ Nonlinear conjugate gradient using curvature
Reimar H Leike's avatar
Reimar H Leike committed
28 29 30 31 32
    The YANGO (Yet Another Nonlinear conjugate Gradient Optimizer)
    uses the curvature to make estimates about suitable descent
    directions. It takes the step that lets it go directly to
    the second order minimum in the subspace spanned by the last
    descent direction and the new gradient.
33 34 35 36 37 38 39 40 41 42 43 44 45 46

    Parameters
    ----------
    controller : IterationController
        Object that decides when to terminate the minimization.

    Notes
    -----
    No restarting procedure has been implemented yet.

    References
    ----------
    """

Martin Reinecke's avatar
Martin Reinecke committed
47 48
    def __init__(self, controller,
                 line_searcher=LineSearchStrongWolfe(c2=0.1)):
49 50 51 52 53 54 55 56 57 58 59 60
        self._controller = controller
        self._line_searcher = line_searcher

    def __call__(self, energy):
        controller = self._controller
        status = controller.start(energy)
        if status != controller.CONTINUE:
            return energy, status
        f_k_minus_1 = None

        p = -energy.gradient
        A_k = energy.curvature
Martin Reinecke's avatar
Martin Reinecke committed
61 62
        energy, success = self._line_searcher.perform_line_search(
            energy, p.vdot(p)/(p.vdot(A_k(p)))*p, f_k_minus_1)
63 64 65
        if not success:
            return energy, controller.ERROR
        A_k = energy.curvature
66
        while True:
Reimar H Leike's avatar
Reimar H Leike committed
67
            r = -energy.gradient
68
            f_k = energy.value
69 70 71 72 73 74
            Ar = A_k(r)
            Ap = A_k(p)
            rAr = r.vdot(Ar)
            pAp = p.vdot(Ap)
            pAr = p.vdot(Ar)
            rAp = r.vdot(Ap)
Reimar H Leike's avatar
Reimar H Leike committed
75 76
            rp = r.vdot(p)
            rr = r.vdot(r)
77
            if rr == 0 or rAr == 0:
Martin Reinecke's avatar
Martin Reinecke committed
78 79
                logger.warning(
                    "Warning: gradient norm 0, assuming convergence!")
80
                return energy, controller.CONVERGED
81
            det = pAp*rAr-np.abs((rAp)*(pAr))
82
            if det < 0:
Martin Reinecke's avatar
Martin Reinecke committed
83 84 85
                logger.error(
                    "Error: negative determinant ({})".format(det))
                return energy, controller.ERROR
86
            if det == 0:
Martin Reinecke's avatar
Martin Reinecke committed
87 88
                # Try 1D Newton Step
                energy, success = self._line_searcher.perform_line_search(
89 90 91
                    energy, rr/rAr*r, f_k_minus_1)
            else:
                a = (rAr*rp - rAp*rr)/det
92
                b = (pAp*rr - pAr*rp)/det
93
                p = a/b*p+r
Martin Reinecke's avatar
Martin Reinecke committed
94
                energy, success = self._line_searcher.perform_line_search(
95
                    energy, p*b, f_k_minus_1)
96 97
            if not success:
                return energy, controller.ERROR
Reimar H Leike's avatar
Reimar H Leike committed
98
            f_k_minus_1 = f_k
99
            status = self._controller.check(energy)
100 101 102
            if status != controller.CONTINUE:
                return energy, status
            A_k = energy.curvature