Best-neighbour-search optimizer. More...
#include <cmtkBestNeighbourOptimizer.h>
Public Types | |
typedef BestNeighbourOptimizer | Self |
This class. | |
typedef Optimizer | Superclass |
Superclass. | |
Public Member Functions | |
BestNeighbourOptimizer (const Self::ParameterType stepFactor=0.5) | |
Constructor. | |
virtual CallbackResult | Optimize (CoordinateVector &, const Self::ParameterType=1, const Self::ParameterType=0) |
Perform the optimization. | |
Private Attributes | |
Self::ParameterType | StepFactor |
Search step factor. |
Best-neighbour-search optimizer.
This class implements a search technique introduced by Studholme et al. By modifying each parameter of the search space by a certain step up- and downwards, all "neighbours" of the current parameter vector are visited. For each of these, the target function (functional) is evaluated. The search then continues from the parameter vector producing the maximum value. If no further improvement is possible, the step size is decreased by a given factor until it reaches a lower bound.
Definition at line 56 of file cmtkBestNeighbourOptimizer.h.
This class.
Reimplemented from cmtk::Optimizer.
Definition at line 62 of file cmtkBestNeighbourOptimizer.h.
Superclass.
Reimplemented from cmtk::Optimizer.
Definition at line 65 of file cmtkBestNeighbourOptimizer.h.
cmtk::BestNeighbourOptimizer::BestNeighbourOptimizer | ( | const Self::ParameterType | stepFactor = 0.5 ) |
[inline] |
Constructor.
Hand functional and callback to parent class and initialize local variables.
stepFactor | Factor by which the search step size is decreased. |
Definition at line 72 of file cmtkBestNeighbourOptimizer.h.
Search step factor.
This variable determines the factor by which to decrease the search step size if no further improvement is possible at a certain resolution. Reasonable values are in the range 0 < StepFactor < 1. For most cases, a value of 0.5 has been found to provide optimum accuracy and performance.
Definition at line 88 of file cmtkBestNeighbourOptimizer.h.