2 Copyright (C) 2008-2018 EDF R&D
4 This file is part of SALOME ADAO module.
6 This library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 This library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with this library; if not, write to the Free Software
18 Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 See http://www.salome-platform.org/ or email : webmaster.salome@opencascade.com
22 Author: Jean-Philippe Argaud, jean-philippe.argaud@edf.fr, EDF R&D
24 .. index:: single: GradientTest
25 .. _section_ref_algorithm_GradientTest:
27 Checking algorithm "*GradientTest*"
28 -----------------------------------
33 This algorithm allows to check the quality of the adjoint operator, by
34 calculating a residue with known theoretical properties. Different residue
35 formula are available.
37 In any cases, one take :math:`\mathbf{dx}_0=Normal(0,\mathbf{x})` and
38 :math:`\mathbf{dx}=\alpha*\mathbf{dx}_0` with :math:`\alpha_0` a user scaling
39 of the initial perturbation, with default to 1. :math:`F` is the calculation
45 One observe the residue coming from the Taylor development of the :math:`F`
46 function, normalized by the value at the nominal point:
48 .. math:: R(\alpha) = \frac{|| F(\mathbf{x}+\alpha*\mathbf{dx}) - F(\mathbf{x}) - \alpha * \nabla_xF(\mathbf{dx}) ||}{|| F(\mathbf{x}) ||}
50 If the residue is decreasing and the decrease change in :math:`\alpha^2` with
51 respect to :math:`\alpha`, it signifies that the gradient is well calculated
52 until the stopping precision of the quadratic decrease, and that :math:`F` is
55 If the residue is decreasing and the decrease change in :math:`\alpha` with
56 respect to :math:`\alpha`, until a certain level after which the residue remains
57 small and constant, it signifies that the :math:`F` is linear and that the
58 residue is decreasing due to the error coming from :math:`\nabla_xF` term
61 "TaylorOnNorm" residue
62 **********************
64 One observe the residue coming from the Taylor development of the :math:`F`
65 function, with respect to the :math:`\alpha` parameter to the square:
67 .. math:: R(\alpha) = \frac{|| F(\mathbf{x}+\alpha*\mathbf{dx}) - F(\mathbf{x}) - \alpha * \nabla_xF(\mathbf{dx}) ||}{\alpha^2}
69 This is a residue essentially similar to the classical Taylor criterion
70 previously described, but its behavior can differ depending on the numerical
71 properties of the calculation.
73 If the residue is constant until a certain level after which the residue will
74 growth, it signifies that the gradient is well calculated until this stopping
75 precision, and that :math:`F` is not linear.
77 If the residue is systematically growing from a very small value with respect to
78 :math:`||F(\mathbf{x})||`, it signifies that :math:`F` is (quasi-)linear and
79 that the gradient calculation is correct until the precision for which the
80 residue reaches the numerical order of :math:`||F(\mathbf{x})||`.
85 One observe the residue based on the gradient approximation:
87 .. math:: R(\alpha) = \frac{|| F(\mathbf{x}+\alpha*\mathbf{dx}) - F(\mathbf{x}) ||}{\alpha}
89 which has to remain stable until the calculation precision is reached.
91 Optional and required commands
92 ++++++++++++++++++++++++++++++
94 .. index:: single: AlgorithmParameters
95 .. index:: single: CheckingPoint
96 .. index:: single: ObservationOperator
97 .. index:: single: AmplitudeOfInitialDirection
98 .. index:: single: EpsilonMinimumExponent
99 .. index:: single: InitialDirection
100 .. index:: single: ResiduFormula
101 .. index:: single: SetSeed
102 .. index:: single: StoreSupplementaryCalculations
104 The general required commands, available in the editing user interface, are the
107 .. include:: snippets/CheckingPoint.rst
109 .. include:: snippets/ObservationOperator.rst
111 The general optional commands, available in the editing user interface, are
112 indicated in :ref:`section_ref_assimilation_keywords`. Moreover, the parameters
113 of the command "*AlgorithmParameters*" allow to choose the specific options,
114 described hereafter, of the algorithm. See
115 :ref:`section_ref_options_Algorithm_Parameters` for the good use of this
118 The options of the algorithm are the following:
120 .. include:: snippets/AmplitudeOfInitialDirection.rst
122 .. include:: snippets/EpsilonMinimumExponent.rst
124 .. include:: snippets/InitialDirection.rst
126 .. include:: snippets/SetSeed.rst
129 .. index:: single: ResiduFormula
131 This key indicates the residue formula that has to be used for the test. The
132 default choice is "Taylor", and the possible ones are "Taylor" (normalized
133 residue of the Taylor development of the operator, which has to decrease
134 with the square power of the perturbation), "TaylorOnNorm" (residue of the
135 Taylor development of the operator with respect to the perturbation to the
136 square, which has to remain constant) and "Norm" (residue obtained by taking
137 the norm of the Taylor development at zero order approximation, which
138 approximate the gradient, and which has to remain constant).
141 ``{"ResiduFormula":"Taylor"}``
143 StoreSupplementaryCalculations
144 .. index:: single: StoreSupplementaryCalculations
146 This list indicates the names of the supplementary variables that can be
147 available at the end of the algorithm. It involves potentially costly
148 calculations or memory consumptions. The default is a void list, none of
149 these variables being calculated and stored by default. The possible names
150 are in the following list: ["CurrentState", "Residu",
151 "SimulatedObservationAtCurrentState"].
154 ``{"StoreSupplementaryCalculations":["CurrentState"]}``
156 Information and variables available at the end of the algorithm
157 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
159 At the output, after executing the algorithm, there are variables and
160 information originating from the calculation. The description of
161 :ref:`section_ref_output_variables` show the way to obtain them by the method
162 named ``get`` of the variable "*ADD*" of the post-processing. The input
163 variables, available to the user at the output in order to facilitate the
164 writing of post-processing procedures, are described in the
165 :ref:`subsection_r_o_v_Inventaire`.
167 The unconditional outputs of the algorithm are the following:
169 .. include:: snippets/Residu.rst
171 The conditional outputs of the algorithm are the following:
173 .. include:: snippets/CurrentState.rst
175 .. include:: snippets/SimulatedObservationAtCurrentState.rst
180 References to other sections:
181 - :ref:`section_ref_algorithm_FunctionTest`
182 - :ref:`section_ref_algorithm_LinearityTest`
183 - :ref:`section_ref_algorithm_TangentTest`
184 - :ref:`section_ref_algorithm_AdjointTest`