import logging
from typing import Dict, Any
from numpy import float64
from scipy.optimize import minimize
from mumott.core.hashing import list_to_hash
from mumott.optimization.loss_functions.base_loss_function import LossFunction
from .base_optimizer import Optimizer
logger = logging.getLogger(__name__)
[docs]class LBFGS(Optimizer):
"""This Optimizer makes the :term:`LBFGS` algorithm from `scipy.optimize
<https://docs.scipy.org/doc/scipy/reference/optimize.html>`_
available for usage with a :class:`LossFunction <module-mumott.optimization.loss_functions>`.
Parameters
----------
loss_function : LossFunction
The :ref:`loss function <loss_functions>` to be minimized using this algorithm.
kwargs : Dict[str, Any]
Miscellaneous options. See notes for valid entries.
Notes
-----
Valid entries in :attr:`kwargs` are
x0
Initial guess for solution vector. Must be the same size as
:attr:`residual_calculator.coefficients`. Defaults to :attr:`loss_function.initial_values`.
bounds
Used to set the ``bounds`` of the optimization method, see `scipy.optimize.minimize
<https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html>`_
documentation for details. Defaults to ``None``.
maxiter
Maximum number of iterations. Defaults to ``10``.
disp
Whether to display output from the optimizer. Defaults to ``False``
maxcor
Maximum number of Hessian corrections to the Jacobian. Defaults to ``10``.
iprint
If ``disp`` is true, controls output with no output if ``iprint < 0``,
convergence output only if ``iprint == 0``, iteration-wise output if
``0 < iprint <= 99``, and sub-iteration output if ``iprint > 99``.
maxfun
Maximum number of function evaluations, including line search evaluations.
Defaults to ``20``.
ftol
Relative change tolerance for objective function. Changes to absolute change tolerance
if objective function is ``< 1``, see `scipy.optimize.minimize
<https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html>`_
documentation, which may lead to excessively fast convergence.
Defaults to ``1e-3``.
gtol
Convergence tolerance for gradient. Defaults to ``1e-5``.
"""
def __init__(self,
loss_function: LossFunction,
**kwargs: Dict[str, Any]):
super().__init__(loss_function, **kwargs)
# This will later be used to reshape the flattened output.
self._output_shape = None
[docs] def optimize(self) -> Dict:
""" Executes the optimization using the options stored in this class
instance. The optimization will continue until convergence,
or until the maximum number of iterations (:attr:`maxiter`) is exceeded.
Returns
-------
A ``dict`` of optimization results. See `scipy.optimize.OptimizeResult
<https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.OptimizeResult.html>`_
for details. The entry ``'x'``, which contains the result, will be reshaped using
the shape of the gradient from :attr:`loss_function`.
"""
lbfgs_kwargs = dict(x0=self._loss_function.initial_values,
bounds=None)
misc_options = dict(maxiter=10,
disp=False,
maxcor=10,
iprint=1,
maxfun=20,
ftol=1e-3,
gtol=1e-5)
for k in lbfgs_kwargs:
if k in dict(self):
lbfgs_kwargs[k] = self[k]
for k in misc_options:
if k in dict(self):
misc_options[k] = self[k]
for k in dict(self):
if k not in lbfgs_kwargs and k not in misc_options:
logger.warning(f'Unknown option {k}, with value {self[k]}, has been ignored.')
lbfgs_kwargs['x0'] = lbfgs_kwargs['x0'].ravel()
with self._tqdm(misc_options['maxiter']) as progress:
def progress_callback(*args, **kwargs):
progress.update(1)
def loss_function_wrapper(coefficients):
d = self._loss_function.get_loss(coefficients, get_gradient=True)
# Store gradient shape to reshape flattened output.
if self._output_shape is None:
self._output_shape = d['gradient'].shape
# LBFGS needs float64
return d['loss'], d['gradient'].ravel().astype(float64)
result = minimize(fun=loss_function_wrapper, callback=progress_callback,
**lbfgs_kwargs, jac=True, method='L-BFGS-B', options=misc_options)
result = dict(result)
result['x'] = result['x'].reshape(self._output_shape)
return dict(result)
def __hash__(self) -> int:
to_hash = [self._options, hash(self._loss_function)]
return int(list_to_hash(to_hash), 16)