BALL::EnergyMinimizer Class Reference
[Energy Minimizer]
#include <BALL/MOLMEC/MINIMIZATION/energyMinimizer.h>
Detailed Description
EnergyMinimizer. Base class for all minimizer for geometry optimization.
Constructor & Destructor Documentation
| BALL::EnergyMinimizer::EnergyMinimizer | ( | ) |
Default constructor.
| BALL::EnergyMinimizer::EnergyMinimizer | ( | ForceField & | force_field | ) |
Constructor.
| BALL::EnergyMinimizer::EnergyMinimizer | ( | ForceField & | force_field, | |
| const Options & | options | |||
| ) |
Constructor.
| BALL::EnergyMinimizer::EnergyMinimizer | ( | const EnergyMinimizer & | energy_minimizer | ) |
Copy constructor
| virtual BALL::EnergyMinimizer::~EnergyMinimizer | ( | ) | [virtual] |
Destructor.
Member Function Documentation
| void BALL::EnergyMinimizer::enableEnergyAbortCondition | ( | bool | state | ) |
Specify if the MDSimulation aborts if the Energy is greater than abort_energy_
| bool BALL::EnergyMinimizer::energyAbortConditionEnabled | ( | ) | const |
Query if the MDSimulation aborts if the Energy is greater than abort_energy_.
| virtual double BALL::EnergyMinimizer::findStep | ( | ) | [virtual] |
Calculate the next step. This method is implemented in each minimizer class and tries to determine the next step to be taken. It typically performs a line search. The value returned is usually the step length with respect to the current direction.
- Returns:
- -1 to indicate failure
Reimplemented in BALL::ConjugateGradientMinimizer, BALL::ShiftedLVMMMinimizer, BALL::SteepestDescentMinimizer, and BALL::StrangLBFGSMinimizer.
| virtual void BALL::EnergyMinimizer::finishIteration | ( | ) | [virtual] |
Finishing step for this iteration. This method should be called at the end of the main iteration loop implemented in minimize . It takes over some administrative stuff:
- increment the iteration counter number_of_iterations_
- call takeSnapShot if necessary
- call printEnergy if necessary
- call ForceField::update if necessary (to rebuild the pair lists!)
- update the same_energy_counter_ tested in isConverged
This method should be overwritten only in rare cases. Even then, the programmer should make sure to call EnergyMinimizer::finishIteration or has to take care of the above items himself.
All derived classes should call this method at the end of the minimize main loop. Otherwise strange things might happen.
- See also:
- minimize
| Gradient& BALL::EnergyMinimizer::getDirection | ( | ) |
Return a reference to the current search direction
| double& BALL::EnergyMinimizer::getEnergy | ( | ) |
Return a reference to the current energy
| double BALL::EnergyMinimizer::getEnergy | ( | ) | const |
Return the current energy
| float BALL::EnergyMinimizer::getEnergyDifferenceBound | ( | ) | const |
Get the energy difference bound
| Size BALL::EnergyMinimizer::getEnergyOutputFrequency | ( | ) | const |
Get the energy ouput frequency
| float BALL::EnergyMinimizer::getEnergyToAbort | ( | ) | const |
| Size BALL::EnergyMinimizer::getEnergyUpdateCounter | ( | ) | const |
Return the number of energy updates since the start of the minimization.
| ForceField* BALL::EnergyMinimizer::getForceField | ( | ) |
Return the force field of the energy minimizer
| Size BALL::EnergyMinimizer::getForceUpdateCounter | ( | ) | const |
Return the number of force updates since the start of the minimization.
| Gradient& BALL::EnergyMinimizer::getGradient | ( | ) |
Return a reference to the current gradient
| double& BALL::EnergyMinimizer::getInitialEnergy | ( | ) |
Return a mutable reference to the initial energy
| double BALL::EnergyMinimizer::getInitialEnergy | ( | ) | const |
Return the initial energy
| Gradient& BALL::EnergyMinimizer::getInitialGradient | ( | ) |
Return a reference to the initial gradient
| float BALL::EnergyMinimizer::getMaxGradient | ( | ) | const |
Get the maximum RMS gradient (first convergence criterion). The gradient unit of the gradient is kJ/(mol ).
| float BALL::EnergyMinimizer::getMaximumDisplacement | ( | ) | const |
Get the maximum displacement value
| Size BALL::EnergyMinimizer::getMaxNumberOfIterations | ( | ) | const |
Get the maximum number of iterations
| Size BALL::EnergyMinimizer::getMaxSameEnergy | ( | ) | const |
Get the maximum number of iterations allowed with equal energy (second convergence criterion)
| Size BALL::EnergyMinimizer::getNumberOfIterations | ( | ) | const |
Return the number of iterations performed.
| Size BALL::EnergyMinimizer::getSnapShotFrequency | ( | ) | const |
Get the snapshot output frequency
| virtual bool BALL::EnergyMinimizer::isConverged | ( | ) | const [virtual] |
Implements the convergence criterion. If the convergence criterion is fulfilled, this method returns true. The convergence criterion is implemented as one of two conditions:
(1) {RMS gradient} is below max_rms_gradient_ (2) same_energy_counter_ is above max_same_energy_
If any of these conditions hold isConverged returns true. This method should be reimplemented in derived classes for a different convergence criterion.
| bool BALL::EnergyMinimizer::isValid | ( | ) | const |
Is the energy minimizer valid : did the setup work?
Minimize the energy of the system bound to the force field. If a number of steps is given, the minimization is aborted after that number of steps, regardless of the number of steps given in the options (MAX_STEPS). Together with the resume option this feature is used to extract properties or visualize the results in the course of the minimization. If resume is set to true, the minimization continues with the former step width and settings.
- Parameters:
-
steps maximum number of steps to be taken resume true if the minimization is to be resumed with the previous settings
- Returns:
- bool - true if the minimization has converged
Reimplemented in BALL::ConjugateGradientMinimizer, BALL::ShiftedLVMMMinimizer, BALL::SteepestDescentMinimizer, and BALL::StrangLBFGSMinimizer.
| const EnergyMinimizer& BALL::EnergyMinimizer::operator= | ( | const EnergyMinimizer & | energy_minimizer | ) |
Assignment operator
| bool BALL::EnergyMinimizer::operator== | ( | const EnergyMinimizer & | energy_minimizer | ) |
Equality operator
| virtual void BALL::EnergyMinimizer::printEnergy | ( | ) | const [virtual] |
Print the energy. This method is called by finishIteration after every energy_output_frequency_ steps. It prints the current RMS gradient and the current energy to Log .info().
| void BALL::EnergyMinimizer::setEnergyDifferenceBound | ( | float | energy_difference_bound | ) |
Set the energy difference bound for convergence
| void BALL::EnergyMinimizer::setEnergyOutputFrequency | ( | Size | energy_output_frequency | ) |
Set the energy output frequency
| void BALL::EnergyMinimizer::setEnergyToAbort | ( | float | value | ) |
Set the value for the energy, that will result in aborting the minization, if it will be surpassed. Default value: 10^9.
| void BALL::EnergyMinimizer::setMaxGradient | ( | float | max_gradient | ) |
Set the maximum RMS gradient (first convergence criterion). The gradient unit of the gradient is kJ/(mol ).
| void BALL::EnergyMinimizer::setMaximumDisplacement | ( | float | maximum_displacement | ) |
Set the maximum displacement value. This is the maximum distance an atom may be moved by the minimizer in one iteration.
| void BALL::EnergyMinimizer::setMaxNumberOfIterations | ( | Size | number_of_iterations | ) |
Set the maximum number of iterations
| void BALL::EnergyMinimizer::setMaxSameEnergy | ( | Size | number | ) |
Set the maximum number of iterations allowed with equal energy (second convergence criterion)
| void BALL::EnergyMinimizer::setNumberOfIterations | ( | Size | number_of_iterations | ) |
Set the number of iterations performed so far.
| void BALL::EnergyMinimizer::setSnapShotFrequency | ( | Size | snapshot_frequency | ) |
Set the snapshot frequency
| bool BALL::EnergyMinimizer::setup | ( | ForceField & | force_field, | |
| const Options & | options | |||
| ) |
Sets up the energy minimizer.
| bool BALL::EnergyMinimizer::setup | ( | ForceField & | force_field, | |
| SnapShotManager * | ssm, | |||
| const Options & | options | |||
| ) |
Sets up the energy minimizer.
| bool BALL::EnergyMinimizer::setup | ( | ForceField & | force_field, | |
| SnapShotManager * | ssm | |||
| ) |
Sets up the energy minimizer.
| bool BALL::EnergyMinimizer::setup | ( | ForceField & | force_field | ) |
Sets up the energy minimizer.
| virtual bool BALL::EnergyMinimizer::specificSetup | ( | ) | [virtual] |
Specific setup
Reimplemented in BALL::ConjugateGradientMinimizer, BALL::ShiftedLVMMMinimizer, BALL::SteepestDescentMinimizer, and BALL::StrangLBFGSMinimizer.
| void BALL::EnergyMinimizer::storeGradientEnergy | ( | ) |
Store the current energy and gradient. The current gradient and current energy is copied into initial energy and initial gradient. This is usually done at the start of an iteration.
| virtual void BALL::EnergyMinimizer::takeSnapShot | ( | ) | const [virtual] |
Take a snapshot of the system. This method is called by finishIteration after every snapshot_frequency_ steps. It saves a SnapShot of the current atom coordinates to a SnapShotManager (if enabled).
- See also:
| virtual void BALL::EnergyMinimizer::updateDirection | ( | ) | [virtual] |
Update the search direction. This method is implemented by the derived classes to implement a method to determine a new search direction.
Reimplemented in BALL::ConjugateGradientMinimizer, BALL::ShiftedLVMMMinimizer, BALL::SteepestDescentMinimizer, and BALL::StrangLBFGSMinimizer.
| virtual double BALL::EnergyMinimizer::updateEnergy | ( | ) | [virtual] |
Update energy. This method calls force_field_->updateEnergy() and stores the result in current_energy_.
| virtual void BALL::EnergyMinimizer::updateForces | ( | ) | [virtual] |
Update forces and store them in current_grad_. This method calls force_field_->updateForces() and stores them in current_grad_ .
| bool BALL::EnergyMinimizer::wasAborted | ( | ) | const |
Return true, if the minimization was aborted, e.g. because of strange energies or gradient.
Member Data Documentation
bool BALL::EnergyMinimizer::abort_by_energy_enabled_ [protected] |
float BALL::EnergyMinimizer::abort_energy_ [protected] |
bool BALL::EnergyMinimizer::aborted_ [protected] |
double BALL::EnergyMinimizer::current_energy_ [protected] |
The current energy.
Gradient BALL::EnergyMinimizer::current_grad_ [protected] |
The current gradient.
float BALL::EnergyMinimizer::cutlo_ [protected] |
Numerical lower bound: we don't want to compute the reciprocal of a number which is lower than 'cutlo_'.
Gradient BALL::EnergyMinimizer::direction_ [protected] |
The current search direction
If the energy difference (before and after an iteration) is smaller than this bound, the minimization procedure stops.
Size BALL::EnergyMinimizer::energy_output_frequency_ [protected] |
Frequency of energy output
Size BALL::EnergyMinimizer::energy_update_counter_ [protected] |
Internal counter: how often is an energy update done. Measure for the speed of minimization.
ForceField* BALL::EnergyMinimizer::force_field_ [protected] |
The force field bound to the energy minimizer. Among other data the force field contains the molecular system whose energy will be minimized by the energy minimizer.
Size BALL::EnergyMinimizer::force_update_counter_ [protected] |
Internal counter: how often is a force update done. Measure for the speed of minimization.
double BALL::EnergyMinimizer::initial_energy_ [protected] |
The energy at the beginning of the current minimization step.
Gradient BALL::EnergyMinimizer::initial_grad_ [protected] |
The gradient at the beginning of the current minimization step.
double BALL::EnergyMinimizer::max_gradient_ [protected] |
The maximum RMS gradient tolerated (first convergence criterion)
Size BALL::EnergyMinimizer::max_same_energy_ [protected] |
The maximum number of iterations with same energy. When this number is reached, we assume the system to have converged (second convergence criterion)
Maximum number of iterations
float BALL::EnergyMinimizer::maximum_displacement_ [protected] |
The maximal shift of an atom per iteration step (in Angstrom).
Size BALL::EnergyMinimizer::number_of_iterations_ [protected] |
The current iteration number
double BALL::EnergyMinimizer::old_energy_ [protected] |
The energy from the last step
Gradient BALL::EnergyMinimizer::old_grad_ [protected] |
The gradient from the last step
Size BALL::EnergyMinimizer::same_energy_counter_ [protected] |
A counter for the number of steps with a similar energy.
SnapShotManager* BALL::EnergyMinimizer::snapshot_ [protected] |
Pointer to a SnapShotManager for storing snapshots of the system
Size BALL::EnergyMinimizer::snapshot_frequency_ [protected] |
Frequency of atom coordinate ouput
double BALL::EnergyMinimizer::step_ [protected] |
The last step size (in respect of the length of the computed direction vector), so the length of the last step was 
.
bool BALL::EnergyMinimizer::valid_ [protected] |
The boolean variable indicates if the setup of the energy minimizer was successful