Package ffx.algorithms.optimize

Class MinimizeOpenMM

java.lang.Object

ffx.algorithms.optimize.Minimize

ffx.algorithms.optimize.MinimizeOpenMM

All Implemented Interfaces:

Terminatable, OptimizationListener

public class MinimizeOpenMM
extends Minimize

Given a Context, this class searches for a new set of particle positions that represent a local minimum of the potential energy. The search is performed with the L-BFGS algorithm. Distance constraints are enforced during minimization by adding a harmonic restraining force to the potential function. The strength of the restraining force is steadily increased until the minimum energy configuration satisfies all constraints to within the tolerance specified by the Context's Integrator.

Energy minimization is done using the force groups defined by the Integrator. If you have called setIntegrationForceGroups() on it to restrict the set of forces used for integration, only the energy of the included forces will be minimized.

Since:

1.0

Author:

Michael J. Schnieders

Nested Class Summary

Nested classes/interfaces inherited from class ffx.algorithms.optimize.Minimize

Minimize.MinimizationEngine

Field Summary

Fields inherited from class ffx.algorithms.optimize.Minimize

algorithmListener, DEFAULT_LBFGS_VECTORS, done, energy, grad, molecularAssembly, n, nSteps, potential, scaling, status, terminate, time, x

Constructor Summary

Constructors

Constructor	Description
MinimizeOpenMM(MolecularAssembly molecularAssembly)	MinimizeOpenMM constructor.
MinimizeOpenMM(MolecularAssembly molecularAssembly, OpenMMEnergy openMMEnergy)	MinimizeOpenMM constructor.
MinimizeOpenMM(MolecularAssembly molecularAssembly, OpenMMEnergy openMMEnergy, AlgorithmListener algorithmListener)	MinimizeOpenMM constructor.

Method Summary

Modifier and Type	Method	Description
Potential	minimize(double eps, int maxIterations)	minimize
Potential	minimize(int m, double eps, int maxIterations)	Note the OpenMM L-BFGS minimizer does not accept the parameter "m" for the number of previous steps used to estimate the Hessian.
boolean	optimizationUpdate(int iteration, int nBFGS, int functionEvaluations, double rmsGradient, double rmsCoordinateChange, double energy, double energyChange, double angle, LineSearch.LineSearchResult lineSearchResult)	MinimizeOpenMM does not support the OptimizationListener interface.

Methods inherited from class ffx.algorithms.optimize.Minimize

defaultEngine, getEnergy, getIterations, getRMSGradient, getStatus, minimize, minimize, minimizeFactory, terminate

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

MinimizeOpenMM

public MinimizeOpenMM(MolecularAssembly molecularAssembly)

MinimizeOpenMM constructor.

Parameters:

molecularAssembly - the MolecularAssembly to optimize.

MinimizeOpenMM

public MinimizeOpenMM(MolecularAssembly molecularAssembly,
 OpenMMEnergy openMMEnergy)

MinimizeOpenMM constructor.

Parameters:

molecularAssembly - the MolecularAssembly to optimize.

openMMEnergy - the OpenMM potential energy function.

MinimizeOpenMM

public MinimizeOpenMM(MolecularAssembly molecularAssembly,
 OpenMMEnergy openMMEnergy,
 AlgorithmListener algorithmListener)

MinimizeOpenMM constructor.

Parameters:

molecularAssembly - the MolecularAssembly to optimize.

openMMEnergy - the OpenMM potential energy function.

algorithmListener - report progress using the listener.

Method Details

minimize

public Potential minimize(int m,
 double eps,
 int maxIterations)

Note the OpenMM L-BFGS minimizer does not accept the parameter "m" for the number of previous steps used to estimate the Hessian.

Overrides:

minimize in class Minimize

Parameters:

m - The number of previous steps used to estimate the Hessian (ignored).

eps - The convergence criteria.

maxIterations - The maximum number of iterations.

Returns:

The potential.

minimize

public Potential minimize(double eps,
 int maxIterations)

minimize

Overrides:

minimize in class Minimize

Parameters:

eps - The convergence criteria.

maxIterations - The maximum number of iterations.

Returns:

a Potential object.

optimizationUpdate

public boolean optimizationUpdate(int iteration,
 int nBFGS,
 int functionEvaluations,
 double rmsGradient,
 double rmsCoordinateChange,
 double energy,
 double energyChange,
 double angle,
 LineSearch.LineSearchResult lineSearchResult)

MinimizeOpenMM does not support the OptimizationListener interface.

Specified by:

optimizationUpdate in interface OptimizationListener

Overrides:

optimizationUpdate in class Minimize

Parameters:

iteration - Number of iterations.

nBFGS - Number of L-BFGS correction vectors.

functionEvaluations - Number of function evaluations.

rmsGradient - RMS gradient at current solution.

rmsCoordinateChange - RMS coordinate change at current solution.

energy - Function value at current solution.

energyChange - Change in the function value compared to the previous solution.

angle - Current angle between gradient and search direction.

lineSearchResult - Result of the line search (null at iteration == 0).

Returns:

A return value of false will terminate the optimization.

Since:

1.0