1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
|
/*=========================================================================
*
* Copyright UMC Utrecht and contributors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0.txt
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
*=========================================================================*/
#ifndef __elxCMAEvolutionStrategy_h
#define __elxCMAEvolutionStrategy_h
#include "elxIncludes.h" // include first to avoid MSVS warning
#include "itkCMAEvolutionStrategyOptimizer.h"
namespace elastix
{
/**
* \class CMAEvolutionStrategy
* \brief An optimizer based on the itk::CMAEvolutionStrategyOptimizer.
*
* A Covariance-Matrix-Adaptation-Evolution-Strategy optimizer.
*
* This optimizer support the NewSamplesEveryIteration option. It requests
* new samples for the computation of each search direction (not during
* the offspring generation). The theory doesn't say anything about such a
* situation, so, think twice before using the NewSamplesEveryIteration option.
*
* The parameters used in this class are:
* \parameter Optimizer: Select this optimizer as follows:\n
* <tt>(Optimizer "CMAEvolutionStrategy")</tt>
* \parameter MaximumNumberOfIterations: The maximum number of iterations in each resolution. \n
* example: <tt>(MaximumNumberOfIterations 100 100 50)</tt> \n
* Default value: 500.\n
* \parameter StepLength: Set the length of the initial step ( = Sigma0 = InitialSigma).\n
* example: <tt>(StepLength 2.0 1.0 0.5)</tt> \n
* Recommended value: 1/3 of the expected parameter range.\n
* Default value: 1.0.\n
* \parameter ValueTolerance: Stopping criterion. See the documentation of the
* itk::CMAEvolutionStrategyOptimizer for more information.\n
* example: <tt>(ValueTolerance 0.001 0.0001 0.000001)</tt> \n
* Default value: 0.00001. Can be specified for each resolution.\n
* \parameter PositionToleranceMin: Stopping criterion. See the documentation of the
* itk::CMAEvolutionStrategyOptimizer for more information.\n
* example: <tt>(PositionToleranceMin 0.001 0.0001 0.000001)</tt> \n
* Default value: 1e-8. Can be specified for each resolution.\n
* \parameter PositionToleranceMax: Stopping criterion. See the documentation of the
* itk::CMAEvolutionStrategyOptimizer for more information.\n
* example: <tt>(PositionToleranceMax 0.001 0.0001 0.000001)</tt> \n
* Default value: 1e8. Can be specified for each resolution.\n
* \parameter PopulationSize: the number of parameter vectors evaluated in each iteration.\n
* If you set it to 0, a default value is calculated by the optimizer, which is reported
* back to the elastix.log file.\n
* example: <tt>(PopulationSize 0 20 20)</tt> \n
* Default: 0 (so, automatically determined). Can be specified for each resolution.\n
* \parameter NumberOfParents: the number of parameter vectors selected for recombination.\n
* If you set it to 0, a default value is calculated
* by the optimizer, which is reported back to the elastix.log file.\n
* example: <tt>(NumberOfParents 0 10 10)</tt> \n
* Default: 0 (so, automatically determined). Can be specified for each resolution. \n
* This value must be less than or equal to the PopulationSize.\n
* \parameter MaximumDeviation: the step length is limited to this value. See the documentation of the
* itk::CMAEvolutionStrategyOptimizer for more information.\n
* example: <tt>(MaximumDeviation 10.0 10.0 5.0)</tt> \n
* Default: 10.0 * positionToleranceMax = practically infinity. Can be specified for each resolution.\n
* \parameter MinimumDeviation: the step length is ensured to be greater than this value.\n
* See the documentation of the itk::CMAEvolutionStrategyOptimizer for more information.\n
* example: <tt>(MinimumDeviation 0.01 0.01 0.0001)</tt> \n
* Default: 0.0. Can be specified for each resolution.\n
* \parameter UseDecayingSigma: use a predefined decaying function to control the steplength sigma.\n
* example: <tt>(UseDecayingSigma "false" "true" "false")</tt> \n
* Default/recommended: "false". Can be specified for each resolution.\n
* If you set it to true the SP_A and SP_alpha parameters apply.\n
* \parameter SP_A: If UseDecayingSigma is set to "true", the steplength \f$sigma(k)\f$ at each
* iteration \f$k\f$ is defined by: \n
* \f$sigma(k+1) = sigma(k) (SP\_A + k)^{SP\_alpha} / (SP\_A + k + 1)^{SP\_alpha}\f$. \n
* where sigma(0) is set by the parameter "StepLength". \n
* example: <tt>(SP_A 50.0 50.0 100.0)</tt> \n
* The default value is 50.0. SP_A can be defined for each resolution. \n
* \parameter SP_alpha: If UseDecayingSigma is set to "true", the steplength \f$sigma(k)\f$ at each
* iteration \f$k\f$ is defined by: \n
* \f$sigma(k+1) = sigma(k) (SP\_A + k)^{SP\_alpha} / (SP\_A + k + 1)^{SP\_alpha}\f$. \n
* where sigma(0) is set by the parameter "StepLength".\n
* example: <tt>(SP_alpha 0.602 0.602 0.602)</tt> \n
* The default value is 0.602. SP_alpha can be defined for each resolution. \n
* \parameter UseCovarianceMatrixAdaptation: a boolean that determines whether to use the
* covariance matrix adaptation scheme.\n
* example: <tt>(UseCovarianceMatrixAdaptation "false" "true" "true")</tt> \n
* Default: "true". This parameter may be altered by the optimizer. The actual value used is \n
* reported back in the elastix.log file. This parameter can be specified for each resolution. \n
* \parameter RecombinationWeightsPreset: the name of a preset for the recombination weights.\n
* See the documentation of the itk::CMAEvolutionStrategyOptimizer for more information.\n
* example: <tt>(UseCovarianceMatrixAdaptation "equal" "linear" "superlinear")</tt> \n
* Default/recommended: "superlinear". Choose one of {"equal", "linear", "superlinear"}.
* This parameter can be specified for each resolution. \n
* \parameter UpdateBDPeriod: the number of iterations after which the eigendecomposition of the
* covariance matrix is updated. If 0, the optimizer estimates a value. The actual value used is
* reported back in the elastix.log file. This parameter can be specified for each resolution. \n
* example: <tt>(UpdateBDPeriod 0 0 50)</tt> \n
* Default: 0 (so, automatically determined).
*
* \ingroup Optimizers
*/
template< class TElastix >
class CMAEvolutionStrategy :
public
itk::CMAEvolutionStrategyOptimizer,
public
OptimizerBase< TElastix >
{
public:
/** Standard ITK.*/
typedef CMAEvolutionStrategy Self;
typedef CMAEvolutionStrategyOptimizer Superclass1;
typedef OptimizerBase< TElastix > Superclass2;
typedef itk::SmartPointer< Self > Pointer;
typedef itk::SmartPointer< const Self > ConstPointer;
/** Method for creation through the object factory. */
itkNewMacro( Self );
/** Run-time type information (and related methods). */
itkTypeMacro( CMAEvolutionStrategy, CMAEvolutionStrategyOptimizer );
/** Name of this class.
* Use this name in the parameter file to select this specific optimizer. \n
* example: <tt>(Optimizer "CMAEvolutionStrategy")</tt>\n
*/
elxClassNameMacro( "CMAEvolutionStrategy" );
/** Typedef's inherited from Superclass1.*/
typedef Superclass1::CostFunctionType CostFunctionType;
typedef Superclass1::CostFunctionPointer CostFunctionPointer;
typedef Superclass1::StopConditionType StopConditionType;
typedef Superclass1::ParametersType ParametersType;
typedef Superclass1::DerivativeType DerivativeType;
typedef Superclass1::ScalesType ScalesType;
/** Typedef's inherited from Elastix.*/
typedef typename Superclass2::ElastixType ElastixType;
typedef typename Superclass2::ElastixPointer ElastixPointer;
typedef typename Superclass2::ConfigurationType ConfigurationType;
typedef typename Superclass2::ConfigurationPointer ConfigurationPointer;
typedef typename Superclass2::RegistrationType RegistrationType;
typedef typename Superclass2::RegistrationPointer RegistrationPointer;
typedef typename Superclass2::ITKBaseType ITKBaseType;
/** Check if any scales are set, and set the UseScales flag on or off;
* after that call the superclass' implementation */
virtual void StartOptimization( void );
/** Methods to set parameters and print output at different stages
* in the registration process.*/
virtual void BeforeRegistration( void );
virtual void BeforeEachResolution( void );
virtual void AfterEachResolution( void );
virtual void AfterEachIteration( void );
virtual void AfterRegistration( void );
protected:
CMAEvolutionStrategy(){}
virtual ~CMAEvolutionStrategy() {}
/** Call the superclass' implementation and print the value of some variables */
virtual void InitializeProgressVariables( void );
private:
CMAEvolutionStrategy( const Self & ); // purposely not implemented
void operator=( const Self & ); // purposely not implemented
};
} // end namespace elastix
#ifndef ITK_MANUAL_INSTANTIATION
#include "elxCMAEvolutionStrategy.hxx"
#endif
#endif // end #ifndef __elxCMAEvolutionStrategy_h
|
