genopt.algorithm.util.gps
Class ModelGPS

java.lang.Object
  genopt.algorithm.Optimizer
      genopt.algorithm.util.gps.ModelGPS

Direct Known Subclasses:

GPSCoordinateSearch

public abstract class ModelGPS
extends Optimizer

Abstract class with model Generalized Pattern Search algorithm.

This project was carried out at:

• Lawrence Berkeley National Laboratory (LBNL), Simulation Research Group,

and supported by

• the U.S. Department of Energy (DOE),
• the Swiss Academy of Engineering Sciences (SATW),
• the Swiss National Energy Fund (NEFF), and
• the Swiss National Science Foundation (SNSF)

GenOpt Copyright (c) 1998-2011, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved.

Version:

GenOpt(R) 3.1.0 (December 8, 2011)

Author:

Michael Wetter

Field Summary

private double	Alpha Exponent used for the control of the mesh size if adaptive precision function evaluations are used
double[][]	basDirMat The base direction matrix.
private Point[]	BesPoi Points for the multi-start algorithms.
protected boolean	checkObjFun A flag whether the cost function value has to be checked for previous matching results.
private double	Delta The mesh size factor
private int	dimF The number of function values
private int	dimX The number of independent variables
static Point[]	EMPTY_SEARCH_SET Array of points with zero elements.
protected static java.lang.String	FAIMOV
private double	fNor The cost function value that will be used in the normalization of the sufficient decrease condition
private int	iniMesSizExp The initial mesh size exponent 's_k'
private int	k The iteration number (counting only the main loop)
private int	mesSizDiv The mesh size divider 'r'
private int	mesSizExpInc The increment of the mesh size exponent 't_k' (fixed for all k)
protected static java.lang.String	NEGDIR
protected int	nMaxSteRed The maximal number of step reduction before the algorithm terminates
private int	nSteRed The number of step reduction done up to now
private int	NumIniPoi Number of intial point of the multi-start algorithm
private FunctionEvaluator	PhiFun Function evaluator for the sufficient decrease condition and the mesh size control, i.e., the function phi
protected static java.lang.String	POSDIR
private java.util.Random	RanGen Random number generator (used for multistart algorithm)
private boolean	RepMin Flag that determines whether the minimum will be reported or not
protected static java.lang.String	SUCMOV
private Point[]	x The sequence of iterates
private double	Zeta Factor that will be multiplied with the function PhiFun for sufficient decrease in cost

Fields inherited from class genopt.algorithm.Optimizer

done, EXCLUDING, FS, INCLUDING, LS, MAINITERATION, ORIGINAL, SUBITERATION, TRANSFORMED

Constructor Summary

ModelGPS(GenOpt genOptData)
Constructor used to run it as a single algorithm.

ModelGPS(int meshSizeDivider, int initialMeshSizeExponent)
Constructor used to run the algorithm in a hybrid algorithm for the last iterations.

Method Summary

private double	_getFunctionValue(FunctionEvaluator f, Point x)
private FunctionEvaluator	_getPrecisionControlFunction(java.lang.String funNam, int include) Makes an instance of the function that is used for a sufficient decrease condition.
private int	_run(Point xIni) Runs the optimization process until a termination criteria is satisfied
protected Point[]	add(Point[] x0, Point[] x1) Adds two vectors of points
void	checkObjectiveFunctionValue(Point x) Checks whether the cost function value has already been obtained previously.
static double[]	getClosestEuclideanMeshPoint(double[] x, double[] xMes, double[] delta) Gets the mesh point, on a rectangular mesh, that is closest to the argument.
protected double	getDelta() Gets the current mesh size factor
Point	getF(Point x) Gets the cost function value and registers it into the data base
protected static int	getIndexLowestFunctionValue(Point[] x)
protected double	getInitialDelta() Gets the initial mesh size divider
protected int	getIterationNumber() Gets the iteration number
protected int	getMaximumNumberOfStepReduction() Gets the maximum number of step reductions.
Point	getMinimumPoint() Gets the minimum point.
protected int	getNumberOfStepReduction() Gets the number of step reductions done up to now.
private Point[]	getRandomPoints(int numPoi) Initializes the iterates for the multi-start algorithm.
protected abstract Point[]	globalSearch(Point[] x, double delta) Abstract method for the global search.
protected boolean	haveSufficientDecrease(Point xNew, Point xBestOld)
private static Point[]	increaseSize(Point[] x)
protected abstract double[][]	initializeBaseDirectionMatrix() Method that initializes the base direction matrix.
protected abstract Point[]	localSearch(Point[] x, double delta) Abstract method for the local search.
private void	parseCommandFile(int meshSizeDivider, int initialMeshSizeExponent) Parses the command file.
void	report(Point x, boolean MainIteration) Reports the new trial
protected void	reportMinimum() Reports the minimum point depending on the flag set by protected void reportMinimum(boolean).
protected void	reportMinimum(boolean doReport) Sets a flag that determines whether the minimum will be reported or not.
int	run(Point x0) Runs the optimization process until a termination criteria is satisfied
protected void	setMaximumNumberOfStepReduction(int maxNumberOfStepReductions) Sets the maximum number of step reductions.

Methods inherited from class genopt.algorithm.Optimizer

algorithmRequiresUsageOfStepNumber, appendToOutputListing, checkMaxIteration, ensureOnlyContinuousParameters, ensureOnlyDiscreteParameters, getAbsAccuracyFunction, getDimensionContinuous, getDimensionDiscrete, getDimensionF, getDimensionX, getDiscreteValueDouble0, getDx, getDx0, getF, getIndex0, getIndex0, getInputValueBoolean, getInputValueDouble, getInputValueDouble, getInputValueInteger, getInputValueInteger, getInputValueString, getInputValueString, getKindOfConstraint, getL, getLengthDiscrete, getMainIterationNumber, getMaximumThreadPoolSize, getMaxIterationNumber, getMode, getObjectiveFunctionName, getOutputPath, getPointerToEqualPoints, getRelAccuracyFunction, getSimulationNumber, getStepNumber, getU, getVariableNameContinuous, getVariableNameDiscrete, getX0, getX0, goToEndOfCommandFile, increaseStepNumber, increaseStepNumber, isFeasible, isNextToken, maxIterationReached, mustStopOptimization, print, println, replace, reportCurrentLowestPoint, resetStepNumber, roundCoordinates, run, setInfo, setMode, setNumberOfMatchingResults, setToFeasibleCoordinate, setToFeasibleCoordinate, setWarning, simulate, throwInputError, useStepNumber, writeStepNumber

Methods inherited from class java.lang.Object

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

Field Detail

EMPTY_SEARCH_SET

public static final Point[] EMPTY_SEARCH_SET

Array of points with zero elements.

dimX

private int dimX

The number of independent variables

dimF

private int dimF

The number of function values

Zeta

private double Zeta

Factor that will be multiplied with the function PhiFun for sufficient decrease in cost

PhiFun

private FunctionEvaluator PhiFun

Function evaluator for the sufficient decrease condition and the mesh size control, i.e., the function phi

Alpha

private double Alpha

Exponent used for the control of the mesh size if adaptive precision function evaluations are used

basDirMat

public double[][] basDirMat

The base direction matrix.
Each column vector is a direction.

mesSizDiv

private int mesSizDiv

The mesh size divider 'r'

iniMesSizExp

private int iniMesSizExp

The initial mesh size exponent 's_k'

mesSizExpInc

private int mesSizExpInc

The increment of the mesh size exponent 't_k' (fixed for all k)

nMaxSteRed

protected int nMaxSteRed

The maximal number of step reduction before the algorithm terminates

nSteRed

private int nSteRed

The number of step reduction done up to now

x

private Point[] x

The sequence of iterates

fNor

private double fNor

The cost function value that will be used in the normalization of the sufficient decrease condition

k

private int k

The iteration number (counting only the main loop)

Delta

private double Delta

The mesh size factor

checkObjFun

protected boolean checkObjFun

A flag whether the cost function value has to be checked for previous matching results.

NumIniPoi

private int NumIniPoi

Number of intial point of the multi-start algorithm

BesPoi

private Point[] BesPoi

Points for the multi-start algorithms. This array holds the best iterate obtained from each starting point.

RepMin

private boolean RepMin

Flag that determines whether the minimum will be reported or not

RanGen

private java.util.Random RanGen

Random number generator (used for multistart algorithm)

SUCMOV

protected static java.lang.String SUCMOV

FAIMOV

protected static java.lang.String FAIMOV

POSDIR

protected static final java.lang.String POSDIR

See Also:

Constant Field Values

NEGDIR

protected static final java.lang.String NEGDIR

See Also:

Constant Field Values

Constructor Detail

ModelGPS

public ModelGPS(GenOpt genOptData)
         throws OptimizerException,
                java.io.IOException,
                java.lang.Exception,
                InputFormatException

Constructor used to run it as a single algorithm.

Parameters:

genOptData - a reference to the GenOpt object.
Note: The object is used as a reference. Hence, its data are modified by this Class.

Throws:

OptimizerException

java.lang.Exception

java.io.IOException - if an I/O exception occurs

InputFormatException

ModelGPS

public ModelGPS(int meshSizeDivider,
                int initialMeshSizeExponent)
         throws OptimizerException,
                java.io.IOException,
                java.lang.Exception,
                InputFormatException

Constructor used to run the algorithm in a hybrid algorithm for the last iterations.

Parameters:

meshSizeDivider - the mesh size divider. If set to Integer.MAX_VALUE, then the value will be read from the command file

initialMeshSizeExponent - the initial mesh size exponent. If set to Integer.MAX_VALUE, then the value will be read from the command file

Throws:

OptimizerException

java.lang.Exception

java.io.IOException - if an I/O exception occurs

InputFormatException

Method Detail

parseCommandFile

private void parseCommandFile(int meshSizeDivider,
                              int initialMeshSizeExponent)
                       throws OptimizerException,
                              java.io.IOException,
                              java.lang.Exception,
                              InputFormatException

Parses the command file.

Parameters:

meshSizeDivider - the mesh size divider. If set to Integer.MAX_VALUE, then the value will be read from the command file

initialMeshSizeExponent - the initial mesh size exponent. If set to Integer.MAX_VALUE, then the value will be read from the command file

Throws:

OptimizerException

java.lang.Exception

java.io.IOException - if an I/O exception occurs

InputFormatException

_getPrecisionControlFunction

private final FunctionEvaluator _getPrecisionControlFunction(java.lang.String funNam,
                                                             int include)
                                                      throws OptimizerException,
                                                             java.io.IOException,
                                                             java.lang.Exception,
                                                             InputFormatException

Makes an instance of the function that is used for a sufficient decrease condition.

Parameters:

funNam - the keyword of the function

include - flag that specifies if zero is includes as a function value or not

Returns:

the function that is used for a sufficient decrease condition

Throws:

OptimizerException

java.lang.Exception

java.io.IOException

InputFormatException

initializeBaseDirectionMatrix

protected abstract double[][] initializeBaseDirectionMatrix()

Method that initializes the base direction matrix.

Each column vector is a direction, and the set of column vectors must be a positive span for the domain of independent parameters.

Returns:

baseDireMat the base direction matrix

getRandomPoints

private Point[] getRandomPoints(int numPoi)

Initializes the iterates for the multi-start algorithm.

Parameters:

numPoi - number of initial points

Returns:

the initial iterates

run

public int run(Point x0)
        throws SimulationInputException,
               OptimizerException,
               java.lang.NoSuchMethodException,
               java.lang.IllegalAccessException,
               java.lang.Exception

Runs the optimization process until a termination criteria is satisfied

Specified by:

run in class Optimizer

Parameters:

x0 - Initial iterate

Returns:

-1 if the maximum number of iteration is exceeded
+1 if the required accuracy is reached

Throws:

OptimizerException - if an OptimizerException occurs or if the user required to stop GenOpt

SimulationInputException - if an error in writing the simulation input file occurs

java.lang.NoSuchMethodException - if a method that should be invoked could not be found

java.lang.IllegalAccessException - if an invoked method enforces Java language access control and the underlying method is inaccessible

java.lang.reflect.InvocationTargetException - if an invoked method throws an exception

java.lang.Exception - if an I/O error in the simulation input file occurs

_run

private int _run(Point xIni)
          throws SimulationInputException,
                 OptimizerException,
                 java.lang.NoSuchMethodException,
                 java.lang.IllegalAccessException,
                 java.lang.Exception

Runs the optimization process until a termination criteria is satisfied

Returns:

-1 if the maximum number of iteration is exceeded
+1 if the required accuracy is reached

Throws:

OptimizerException - if an OptimizerException occurs or if the user required to stop GenOpt

SimulationInputException - if an error in writing the simulation input file occurs

java.lang.NoSuchMethodException - if a method that should be invoked could not be found

java.lang.IllegalAccessException - if an invoked method enforces Java language access control and the underlying method is inaccessible

java.lang.reflect.InvocationTargetException - if an invoked method throws an exception

java.lang.Exception - if an I/O error in the simulation input file occurs

_getFunctionValue

private final double _getFunctionValue(FunctionEvaluator f,
                                       Point x)
                                throws OptimizerException,
                                       java.lang.NoSuchMethodException,
                                       java.lang.IllegalAccessException,
                                       java.lang.reflect.InvocationTargetException

Throws:

OptimizerException

java.lang.NoSuchMethodException

java.lang.IllegalAccessException

java.lang.reflect.InvocationTargetException

haveSufficientDecrease

protected final boolean haveSufficientDecrease(Point xNew,
                                               Point xBestOld)
                                        throws OptimizerException,
                                               java.lang.NoSuchMethodException,
                                               java.lang.IllegalAccessException,
                                               java.lang.reflect.InvocationTargetException

Throws:

OptimizerException

java.lang.NoSuchMethodException

java.lang.IllegalAccessException

java.lang.reflect.InvocationTargetException

add

protected final Point[] add(Point[] x0,
                            Point[] x1)

Adds two vectors of points

Parameters:

x0 - first vector of points

x1 - second vector of points

Returns:

the new vector of points

getInitialDelta

protected final double getInitialDelta()

Gets the initial mesh size divider

Returns:

the initial mesh size divider

getDelta

protected final double getDelta()

Gets the current mesh size factor

Returns:

the current mesh size factor

getIterationNumber

protected final int getIterationNumber()

Gets the iteration number

Returns:

the iteration number

increaseSize

private static final Point[] increaseSize(Point[] x)

getIndexLowestFunctionValue

protected static int getIndexLowestFunctionValue(Point[] x)

reportMinimum

protected void reportMinimum(boolean doReport)

Sets a flag that determines whether the minimum will be reported or not. This is used by algorithms that run ModelGPS multiple times, such as multi-start methods, and need to prevent it from reporting the minimum.

Parameters:

doReport -

reportMinimum

protected void reportMinimum()
                      throws java.io.IOException

Reports the minimum point depending on the flag set by protected void reportMinimum(boolean).
This method gets the minimum point from the data base, calls the function report with an corresponding comment and reports the minimum point to the output device.

Overrides:

reportMinimum in class Optimizer

Throws:

java.io.IOException - if an I/O error in the optimization output files occurs

getMinimumPoint

public Point getMinimumPoint()

Gets the minimum point.

Overrides:

getMinimumPoint in class Optimizer

Returns:

the point with the lowest function value

globalSearch

protected abstract Point[] globalSearch(Point[] x,
                                        double delta)
                                 throws SimulationInputException,
                                        OptimizerException,
                                        java.lang.NoSuchMethodException,
                                        java.lang.IllegalAccessException,
                                        java.lang.Exception

Abstract method for the global search.

Parameters:

x - Sequence of previous iterates

delta - current mesh size

Returns:

the sequence of points that have been evaluated in the global search. Each point must be a mesh point. If no point has been evaluated, then this method must return EMPTY_SEARCH_SET.

Throws:

OptimizerException - if an OptimizerException occurs or if the user required to stop GenOpt

SimulationInputException - if an error in writing the simulation input file occurs

java.lang.NoSuchMethodException - if a method that should be invoked could not be found

java.lang.IllegalAccessException - if an invoked method enforces Java language access control and the underlying method is inaccessible

java.lang.reflect.InvocationTargetException - if an invoked method throws an exception

java.lang.Exception - if an I/O error in the simulation input file occurs

localSearch

protected abstract Point[] localSearch(Point[] x,
                                       double delta)
                                throws SimulationInputException,
                                       OptimizerException,
                                       java.lang.NoSuchMethodException,
                                       java.lang.IllegalAccessException,
                                       java.lang.Exception

Abstract method for the local search.

Parameters:

x - Sequence of previous iterates

delta - current mesh size

Returns:

the sequence of points that have been evaluated in the local search, regardless whether they reduced the cost function value or not.

Throws:

OptimizerException - if an OptimizerException occurs or if the user required to stop GenOpt

SimulationInputException - if an error in writing the simulation input file occurs

java.lang.NoSuchMethodException - if a method that should be invoked could not be found

java.lang.IllegalAccessException - if an invoked method enforces Java language access control and the underlying method is inaccessible

java.lang.reflect.InvocationTargetException - if an invoked method throws an exception

java.lang.Exception - if an I/O error in the simulation input file occurs

checkObjectiveFunctionValue

public void checkObjectiveFunctionValue(Point x)
                                 throws OptimizerException

Checks whether the cost function value has already been obtained previously.
If it has been obtained previously, an information message is reported. If the maximum number of matching function value is obtained, an exception is thrown.

Overrides:

checkObjectiveFunctionValue in class Optimizer

Parameters:

x - the point to be checked

Throws:

OptimizerException - thrown if the maximum number of matching function value is obtained

report

public void report(Point x,
                   boolean MainIteration)
            throws java.io.IOException

Reports the new trial

Overrides:

report in class Optimizer

Parameters:

x - the point to be reported

MainIteration - true if step was a main iteration or false if it was a sub iteration

Throws:

java.io.IOException - if an I/O error in the optimization output files occurs

getF

public Point getF(Point x)
           throws SimulationInputException,
                  OptimizerException,
                  java.lang.NoSuchMethodException,
                  java.lang.IllegalAccessException,
                  java.lang.Exception

Gets the cost function value and registers it into the data base

Overrides:

getF in class Optimizer

Parameters:

x - the point being evaluated

Returns:

a clone of the point with the new function values stored

Throws:

OptimizerException - if an OptimizerException occurs or if the user required to stop GenOpt

SimulationInputException - if an error in writing the simulation input file occurs

java.lang.NoSuchMethodException - if a method that should be invoked could not be found

java.lang.IllegalAccessException - if an invoked method enforces Java language access control and the underlying method is inaccessible

java.lang.reflect.InvocationTargetException - if an invoked method throws an exception

java.lang.Exception - if an I/O error in the simulation input file occurs

getNumberOfStepReduction

protected int getNumberOfStepReduction()

Gets the number of step reductions done up to now. The return value is a zero-based counter.

Returns:

the number of step reductions done up to now

getMaximumNumberOfStepReduction

protected int getMaximumNumberOfStepReduction()

Gets the maximum number of step reductions.

Returns:

the maximum number of step reductions

setMaximumNumberOfStepReduction

protected void setMaximumNumberOfStepReduction(int maxNumberOfStepReductions)

Sets the maximum number of step reductions.

Parameters:

maxNumberOfStepReductions - the maximum number of step reductions

getClosestEuclideanMeshPoint

public static double[] getClosestEuclideanMeshPoint(double[] x,
                                                    double[] xMes,
                                                    double[] delta)

Gets the mesh point, on a rectangular mesh, that is closest to the argument.

Parameters:

x - independent parameter

xMes - any point on the mesh

delta - mesh size factor times the step size of the variable

Returns:

the mesh point that is closest to x in the Euclidean norm.