ITK
6.0.0
Insight Toolkit
|
#include <itkParticleSwarmOptimizerBase.h>
Abstract implementation of a Particle Swarm Optimization (PSO) algorithm.
The PSO algorithm was originally presented in:
J. Kennedy, R. Eberhart, "Particle Swarm Optimization", Proc. IEEE Int. Neural Networks, 1995.
The algorithm is a stochastic global search optimization approach. Optimization is performed by maintaining a swarm of particles that traverse the parameter space, searching for the optimal function value. Associated with each particle are its location and speed, in parameter space.
Swarm initialization is performed within the user supplied parameter bounds using either a uniform distribution or a normal distribution centered on the initial parameter values supplied by the user. The search terminates when the maximal number of iterations has been reached or when the change in the best value in the past \(g\) generations is below a threshold and the swarm has collapsed (i.e. best personal particle locations are close to the swarm's best location in parameter space).
The actual optimization procedure, updating the swarm, is performed in the subclasses, required to implement the UpdateSwarm() method.
NOTE: This implementation only performs minimization.
Definition at line 56 of file itkParticleSwarmOptimizerBase.h.
Classes | |
struct | ParticleData |
Protected Member Functions | |
void | FileInitialization () |
virtual void | Initialize () |
ParticleSwarmOptimizerBase () | |
void | PrintParamtersType (const ParametersType &x, std::ostream &os) const |
void | PrintSelf (std::ostream &os, Indent indent) const override |
void | RandomInitialization () |
virtual void | UpdateSwarm ()=0 |
virtual void | ValidateSettings () |
~ParticleSwarmOptimizerBase () override | |
Protected Member Functions inherited from itk::SingleValuedNonLinearOptimizer | |
void | PrintSelf (std::ostream &os, Indent indent) const override |
SingleValuedNonLinearOptimizer () | |
~SingleValuedNonLinearOptimizer () override=default | |
Protected Member Functions inherited from itk::NonLinearOptimizer | |
NonLinearOptimizer ()=default | |
~NonLinearOptimizer () override | |
Protected Member Functions inherited from itk::Optimizer | |
Optimizer () | |
void | PrintSelf (std::ostream &os, Indent indent) const override |
virtual void | SetCurrentPosition (const ParametersType ¶m) |
~Optimizer () override=default | |
Protected Member Functions inherited from itk::Object | |
Object () | |
bool | PrintObservers (std::ostream &os, Indent indent) const |
void | PrintSelf (std::ostream &os, Indent indent) const override |
virtual void | SetTimeStamp (const TimeStamp &timeStamp) |
~Object () override | |
Protected Member Functions inherited from itk::LightObject | |
virtual LightObject::Pointer | InternalClone () const |
LightObject () | |
virtual void | PrintHeader (std::ostream &os, Indent indent) const |
virtual void | PrintSelf (std::ostream &os, Indent indent) const |
virtual void | PrintTrailer (std::ostream &os, Indent indent) const |
virtual | ~LightObject () |
Additional Inherited Members | |
Static Public Member Functions inherited from itk::SingleValuedNonLinearOptimizer | |
static Pointer | New () |
Static Public Member Functions inherited from itk::NonLinearOptimizer | |
static Pointer | New () |
Static Public Member Functions inherited from itk::Optimizer | |
static Pointer | New () |
Static Public Member Functions inherited from itk::Object | |
static bool | GetGlobalWarningDisplay () |
static void | GlobalWarningDisplayOff () |
static void | GlobalWarningDisplayOn () |
static Pointer | New () |
static void | SetGlobalWarningDisplay (bool val) |
Static Public Member Functions inherited from itk::LightObject | |
static void | BreakOnError () |
static Pointer | New () |
using itk::ParticleSwarmOptimizerBase::ConstPointer = SmartPointer<const Self> |
Definition at line 65 of file itkParticleSwarmOptimizerBase.h.
Definition at line 85 of file itkParticleSwarmOptimizerBase.h.
using itk::ParticleSwarmOptimizerBase::NumberOfGenerationsType = unsigned int |
Definition at line 84 of file itkParticleSwarmOptimizerBase.h.
using itk::ParticleSwarmOptimizerBase::NumberOfIterationsType = unsigned int |
Definition at line 82 of file itkParticleSwarmOptimizerBase.h.
using itk::ParticleSwarmOptimizerBase::NumberOfParticlesType = unsigned int |
Definition at line 83 of file itkParticleSwarmOptimizerBase.h.
using itk::ParticleSwarmOptimizerBase::ParameterBoundsType = std::vector<std::pair<ParametersType::ValueType, ParametersType::ValueType> > |
Definition at line 70 of file itkParticleSwarmOptimizerBase.h.
Definition at line 64 of file itkParticleSwarmOptimizerBase.h.
using itk::ParticleSwarmOptimizerBase::RandomVariateGeneratorType = Statistics::MersenneTwisterRandomVariateGenerator |
Definition at line 87 of file itkParticleSwarmOptimizerBase.h.
Standard "Self" type alias.
Definition at line 62 of file itkParticleSwarmOptimizerBase.h.
Definition at line 63 of file itkParticleSwarmOptimizerBase.h.
using itk::ParticleSwarmOptimizerBase::SwarmType = std::vector<ParticleData> |
Definition at line 81 of file itkParticleSwarmOptimizerBase.h.
Definition at line 86 of file itkParticleSwarmOptimizerBase.h.
|
protected |
|
overrideprotected |
void itk::ParticleSwarmOptimizerBase::ClearSwarm | ( | ) |
Specify the initial swarm. Useful for evaluating PSO variants. If the initial swarm is set it will be used. To revert to random initialization (uniform or normal particle distributions) set using an empty swarm.
|
protected |
|
virtual |
|
virtual |
Specify whether to initialize the particles using a normal distribution centered on the user supplied initial value or a uniform distribution. If the optimum is expected to be near the initial value it is likely that initializing with a normal distribution will result in faster convergence.
|
virtual |
Set/Get maximal number of iterations - the maximal number of function evaluations is m_MaximalNumberOfIterations*m_NumberOfParticles
|
overridevirtual |
Reimplemented from itk::NonLinearOptimizer.
|
virtual |
Set/Get the number of generations to continue with minimal improvement in the function value, |f_best(g_i) - f_best(g_k)|<threshold where k <= i+NumberOfGenerationsWithMinimalImprovement Minimal value is one.
|
virtual |
Set/Get number of particles in the swarm - the maximal number of function evaluations is m_MaximalNumberOfIterations*m_NumberOfParticles
ParameterBoundsType itk::ParticleSwarmOptimizerBase::GetParameterBounds | ( | ) | const |
|
virtual |
Set parameters convergence tolerance using the same value for all, sz, parameters
|
virtual |
Set parameters convergence tolerance using the same value for all, sz, parameters
|
virtual |
Indicate whether or not to output the swarm information when printing an object. By default this option is turned off as it generates too much information.
|
virtual |
Set the random number seed for the swarm. Use this method to produce repeatable results, typically, for testing.
|
overridevirtual |
Get the reason for termination
Reimplemented from itk::Optimizer.
|
virtual |
Use a specific seed to initialize the random number generator. If On, use m_Seed to seed the random number generator. Default is Off.
MeasureType itk::ParticleSwarmOptimizerBase::GetValue | ( | ) | const |
Get the function value for the current position. NOTE: This value is only valid during and after the execution of the StartOptimization() method.
|
protectedvirtual |
Initialize the particle swarm, and seed the random number generator.
|
virtual |
Specify whether to initialize the particles using a normal distribution centered on the user supplied initial value or a uniform distribution. If the optimum is expected to be near the initial value it is likely that initializing with a normal distribution will result in faster convergence.
|
protected |
|
overrideprotectedvirtual |
Methods invoked by Print() to print information about the object including superclasses. Typically not called by the user (use Print() instead) but used in the hierarchical print process to combine the output of several classes.
Reimplemented from itk::Optimizer.
void itk::ParticleSwarmOptimizerBase::PrintSwarm | ( | std::ostream & | os, |
Indent | indent | ||
) | const |
Print the swarm information to the given output stream. Each line (particle data) is of the form: current_parameters current_velocity current_value best_parameters best_value
|
virtual |
Indicate whether or not to output the swarm information when printing an object. By default this option is turned off as it generates too much information.
|
protected |
|
virtual |
The optimization algorithm will terminate when the function improvement in the last m_NumberOfGenerationsWithMinimalImprovement generations is less than m_FunctionConvergenceTolerance and the maximal distance between particles and the best particle in each dimension is less than m_ParametersConvergenceTolerance[i] for the specified percentage of the particles. That is, we haven't improved the best function value for a while and in the parameter space most (m%) of our particles attained their best value close to the swarm's best value. NOTE: The use of different tolerances for each dimension is desired when optimizing over non-commensurate parameters (e.g. rotation and translation). Alternatively, we could use ITK's parameter scaling approach. The current approach seems more intuitive.
|
virtual |
Specify whether to initialize the particles using a normal distribution centered on the user supplied initial value or a uniform distribution. If the optimum is expected to be near the initial value it is likely that initializing with a normal distribution will result in faster convergence.
void itk::ParticleSwarmOptimizerBase::SetInitialSwarm | ( | const SwarmType & | initialSwarm | ) |
Specify the initial swarm. Useful for evaluating PSO variants. If the initial swarm is set it will be used. To revert to random initialization (uniform or normal particle distributions) set using an empty swarm.
|
virtual |
Set/Get maximal number of iterations - the maximal number of function evaluations is m_MaximalNumberOfIterations*m_NumberOfParticles
|
virtual |
Set/Get the number of generations to continue with minimal improvement in the function value, |f_best(g_i) - f_best(g_k)|<threshold where k <= i+NumberOfGenerationsWithMinimalImprovement Minimal value is one.
void itk::ParticleSwarmOptimizerBase::SetNumberOfParticles | ( | NumberOfParticlesType | n | ) |
Set/Get number of particles in the swarm - the maximal number of function evaluations is m_MaximalNumberOfIterations*m_NumberOfParticles
|
virtual |
Set/Get the parameter bounds. Search for optimal value is inside these bounds. NOTE: It is assumed that the first entry is the minimal value, second is the maximal value.
void itk::ParticleSwarmOptimizerBase::SetParameterBounds | ( | std::pair< ParametersType::ValueType, ParametersType::ValueType > & | bounds, |
unsigned int | n | ||
) |
Set/Get the parameter bounds. Search for optimal value is inside these bounds. NOTE: It is assumed that the first entry is the minimal value, second is the maximal value.
|
virtual |
Set parameters convergence tolerance using the same value for all, sz, parameters
void itk::ParticleSwarmOptimizerBase::SetParametersConvergenceTolerance | ( | ValueType | convergenceTolerance, |
unsigned int | sz | ||
) |
Set parameters convergence tolerance using the same value for all, sz, parameters
|
virtual |
Set parameters convergence tolerance using the same value for all, sz, parameters
|
virtual |
Indicate whether or not to output the swarm information when printing an object. By default this option is turned off as it generates too much information.
|
virtual |
Set the random number seed for the swarm. Use this method to produce repeatable results, typically, for testing.
|
virtual |
Use a specific seed to initialize the random number generator. If On, use m_Seed to seed the random number generator. Default is Off.
|
overridevirtual |
Start optimization.
Reimplemented from itk::Optimizer.
|
protectedpure virtual |
Implement your update rule in this function.
Implemented in itk::InitializationBiasedParticleSwarmOptimizer, and itk::ParticleSwarmOptimizer.
|
virtual |
Use a specific seed to initialize the random number generator. If On, use m_Seed to seed the random number generator. Default is Off.
|
protectedvirtual |
|
protected |
Definition at line 255 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 256 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 253 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 246 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 258 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 248 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 249 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 247 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 250 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 257 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 251 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 254 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 252 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 244 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 259 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 245 of file itkParticleSwarmOptimizerBase.h.
|
protected |
Definition at line 260 of file itkParticleSwarmOptimizerBase.h.