← Back
Editing: _direct_py.cpython-311.pyc
� d�cT. � � � d dl mZ d dlmZmZmZmZmZmZm Z d dl Z d dl m Z ddlmZmZ ddlmZ er d dlmZ d dlmZ d gZd Zd Zd d ddde j d dddd� d*d)�ZdS )+� )� annotations)�Any�Callable�Iterable�Optional�Tuple� TYPE_CHECKING�UnionN)�OptimizeResult� )�old_bound_to_new�Bounds)�direct)�NoneTyper ) z<Number of function evaluations done is larger than maxfun={}z.Number of iterations is larger than maxiter={}zu[i] < l[i] for some izmaxfun is too largezInitialization failedz7There was an error in the creation of the sample pointsz/An error occured while the function was sampledz*Maximum number of levels has been reached.z Forced stopzInvalid argumentsz Out of memory)z_The best function value found is within a relative error={} of the (known) global optimum f_minz_The volume of the hyperrectangle containing the lowest function value found is below vol_tol={}zlThe side length measure of the hyperrectangle containing the lowest function value found is below len_tol={}� g-C ��6?�� Tg��ؗ�Ҝ<g���ư>) �args�eps�maxfun�maxiter�locally_biased�f_min� f_min_rtol�vol_tol�len_tol�callback�func�,Callable[[npt.ArrayLike, Tuple[Any]], float]�bounds�Union[Iterable, Bounds]r �tupler �floatr �Union[int, None]r �intr �boolr r r r r �-Optional[Callable[[npt.ArrayLike], NoneType]]�returnr c �� � � t |t � � s^t |t � � st |t � � r#t |� � \ } } t | | � � }nd}t |� � �t j |j t j �� � } t j |j t j �� � } t j | | k � � st d� � �t j t j | � � � � s&t j t j | � � � � rt d� � �| dk s| dk rt d� � �| dk s| dk rt d� � �|dk s|dk rt d � � �|�d | j d z }t |t � � st d � � �|dk rt d � � �t |t � � st d� � �|dk rt d� � �t |t � � st d� � �d� fd� }t# |t j | � � t j | � � |d||||||| | | � � \ }}}}}|||| | f}|dk r-t& |dz � ||dz � � }n{d|cxk rdk r0n n-t* |dz � ||dz � � }n>d|cxk rdk r!n n t* t- |� � dz }nt* |dz }t/ t j |� � |||dk |||�� � S )a� Finds the global minimum of a function using the DIRECT algorithm. Parameters ---------- func : callable The objective function to be minimized. ``func(x, *args) -> float`` where ``x`` is an 1-D array with shape (n,) and ``args`` is a tuple of the fixed parameters needed to completely specify the function. bounds : sequence or `Bounds` Bounds for variables. There are two ways to specify the bounds: 1. Instance of `Bounds` class. 2. ``(min, max)`` pairs for each element in ``x``. args : tuple, optional Any additional fixed parameters needed to completely specify the objective function. eps : float, optional Minimal required difference of the objective function values between the current best hyperrectangle and the next potentially optimal hyperrectangle to be divided. In consequence, `eps` serves as a tradeoff between local and global search: the smaller, the more local the search becomes. Default is 1e-4. maxfun : int or None, optional Approximate upper bound on objective function evaluations. If `None`, will be automatically set to ``1000 * N`` where ``N`` represents the number of dimensions. Will be capped if necessary to limit DIRECT's RAM usage to app. 1GiB. This will only occur for very high dimensional problems and excessive `max_fun`. Default is `None`. maxiter : int, optional Maximum number of iterations. Default is 1000. locally_biased : bool, optional If `True` (default), use the locally biased variant of the algorithm known as DIRECT_L. If `False`, use the original unbiased DIRECT algorithm. For hard problems with many local minima, `False` is recommended. f_min : float, optional Function value of the global optimum. Set this value only if the global optimum is known. Default is ``-np.inf``, so that this termination criterion is deactivated. f_min_rtol : float, optional Terminate the optimization once the relative error between the current best minimum `f` and the supplied global minimum `f_min` is smaller than `f_min_rtol`. This parameter is only used if `f_min` is also set. Must lie between 0 and 1. Default is 1e-4. vol_tol : float, optional Terminate the optimization once the volume of the hyperrectangle containing the lowest function value is smaller than `vol_tol` of the complete search space. Must lie between 0 and 1. Default is 1e-16. len_tol : float, optional If `locally_biased=True`, terminate the optimization once half of the normalized maximal side length of the hyperrectangle containing the lowest function value is smaller than `len_tol`. If `locally_biased=False`, terminate the optimization once half of the normalized diagonal of the hyperrectangle containing the lowest function value is smaller than `len_tol`. Must lie between 0 and 1. Default is 1e-6. callback : callable, optional A callback function with signature ``callback(xk)`` where ``xk`` represents the best function value found so far. Returns ------- res : OptimizeResult The optimization result represented as a ``OptimizeResult`` object. Important attributes are: ``x`` the solution array, ``success`` a Boolean flag indicating if the optimizer exited successfully and ``message`` which describes the cause of the termination. See `OptimizeResult` for a description of other attributes. Notes ----- DIviding RECTangles (DIRECT) is a deterministic global optimization algorithm capable of minimizing a black box function with its variables subject to lower and upper bound constraints by sampling potential solutions in the search space [1]_. The algorithm starts by normalising the search space to an n-dimensional unit hypercube. It samples the function at the center of this hypercube and at 2n (n is the number of variables) more points, 2 in each coordinate direction. Using these function values, DIRECT then divides the domain into hyperrectangles, each having exactly one of the sampling points as its center. In each iteration, DIRECT chooses, using the `eps` parameter which defaults to 1e-4, some of the existing hyperrectangles to be further divided. This division process continues until either the maximum number of iterations or maximum function evaluations allowed are exceeded, or the hyperrectangle containing the minimal value found so far becomes small enough. If `f_min` is specified, the optimization will stop once this function value is reached within a relative tolerance. The locally biased variant of DIRECT (originally called DIRECT_L) [2]_ is used by default. It makes the search more locally biased and more efficient for cases with only a few local minima. A note about termination criteria: `vol_tol` refers to the volume of the hyperrectangle containing the lowest function value found so far. This volume decreases exponentially with increasing dimensionality of the problem. Therefore `vol_tol` should be decreased to avoid premature termination of the algorithm for higher dimensions. This does not hold for `len_tol`: it refers either to half of the maximal side length (for ``locally_biased=True``) or half of the diagonal of the hyperrectangle (for ``locally_biased=False``). This code is based on the DIRECT 2.0.4 Fortran code by Gablonsky et al. at https://ctk.math.ncsu.edu/SOFTWARE/DIRECTv204.tar.gz . This original version was initially converted via f2c and then cleaned up and reorganized by Steven G. Johnson, August 2007, for the NLopt project. The `direct` function wraps the C implementation. .. versionadded:: 1.9.0 References ---------- .. [1] Jones, D.R., Perttunen, C.D. & Stuckman, B.E. Lipschitzian optimization without the Lipschitz constant. J Optim Theory Appl 79, 157-181 (1993). .. [2] Gablonsky, J., Kelley, C. A Locally-Biased form of the DIRECT Algorithm. Journal of Global Optimization 21, 27-37 (2001). Examples -------- The following example is a 2-D problem with four local minima: minimizing the Styblinski-Tang function (https://en.wikipedia.org/wiki/Test_functions_for_optimization). >>> from scipy.optimize import direct, Bounds >>> def styblinski_tang(pos): ... x, y = pos ... return 0.5 * (x**4 - 16*x**2 + 5*x + y**4 - 16*y**2 + 5*y) >>> bounds = Bounds([-4., -4.], [4., 4.]) >>> result = direct(styblinski_tang, bounds) >>> result.x, result.fun, result.nfev array([-2.90321597, -2.90321597]), -78.3323279095383, 2011 The correct global minimum was found but with a huge number of function evaluations (2011). Loosening the termination tolerances `vol_tol` and `len_tol` can be used to stop DIRECT earlier. >>> result = direct(styblinski_tang, bounds, len_tol=1e-3) >>> result.x, result.fun, result.nfev array([-2.9044353, -2.9044353]), -78.33230330754142, 207 z5bounds must be a sequence or instance of Bounds class)�dtypez#Bounds are not consistent min < maxzBounds must not be inf.r r z vol_tol must be between 0 and 1.z len_tol must be between 0 and 1.z#f_min_rtol must be between 0 and 1.Nr zmaxfun must be of type int.zmaxfun must be > 0.z maxiter must be of type int.zmaxiter must be > 0.z%locally_biased must be True or False.c � �� t j | � � } |� �| � � }n �| g|�R � }t j |� � � � � S �N)�np�asarray�item)�xr �fr s ��;/usr/lib/python3/dist-packages/scipy/optimize/_direct_py.py� _func_wrapzdirect.<locals>._func_wrap� sQ �� � �J�q�M�M�� �<���Q���A�A���Q������A��z�!�}�}�!�!�#�#�#� F� � i�����c )r/ �fun�status�success�message�nfev�nitr+ )� isinstancer �listr! r � ValueErrorr, �ascontiguousarray�lb�float64�ub�all�any�isinf�shaper$ r% �_directr- �SUCCESS_MESSAGES�format�ERROR_MESSAGES�absr )r r r r r r r r r r r r rA rC r: r2 r/ r7 �ret_coder; r< � format_vals ` r1 r r ) s� �� �B �f�f� %� %� &� �f�d� #� #� &�z�&�%�'@�'@� &�%�f�-�-�F�B���B��^�^�F�F�2�G� �W�%�%� %� � �f�i�r�z� :� :� :�B� � �f�i�r�z� :� :� :�B� �6�"�r�'�?�?� @��>�?�?�?� ��r�x��|�|� � � 4���r�x��|�|� 4� 4� 4��2�3�3�3� �!� � �w��{�{��;�<�<�<��!� � �w��{�{��;�<�<�<��Q���*�q�.�.��>�?�?�?� �~����� �#�� �f�c� "� "� 8��6�7�7�7� ��z�z��.�/�/�/� �g�s� #� #� 9��7�8�8�8���{�{��/�0�0�0� �n�d� +� +� B��@�A�A�A�$� $� $� $� $� $� #*�� � �2���� �2��� � �s�F�G�� �z���(�#� #� �A�s�H�d�C� �'�:�w��@�J��!�|�|�"�8�a�<�0�7�7� �x�!�|�,�.� .��� �X� � � � �� � � � � � ��A� �.�5�5�j��A� �6N�O�O��� �X� � � � �� � � � � � ��X� � ��!2�3��� ��B� �/�� �B�J�q�M�M�s�8�"*�Q�,�� #�� .� .� .� .r3 )r r r r r r! r r" r r# r r$ r r% r r" r r" r r" r r" r r& r' r )� __future__r �typingr r r r r r r �numpyr, �scipy.optimizer � _constraintsr r rH r � numpy.typing�npt� _typeshedr �__all__rK rI �infr r3 r1 �<module>rY sQ �� "� "� "� "� "� "�� � � � � � � � � � � � � � � � � � � � � � )� )� )� )� )� )� 2� 2� 2� 2� 2� 2� 2� 2� &� &� &� &� &� &�� #� � � � � � �"�"�"�"�"�"� �*�� �� � � �� #�� ��F�7� ���>B�n.� n.� n.� n.� n.� n.� n.� n.r3
Save File
Cancel