edu.stanford.rsl.jpop.fortran
Class UncminForJava

java.lang.Object
  edu.stanford.rsl.jpop.fortran.UncminForJava

public class UncminForJava
extends java.lang.Object

This class contains Java translations of the UNCMIN unconstrained optimization routines. See R.B. Schnabel, J.E. Koontz, and B.E. Weiss, A Modular System of Algorithms for Unconstrained Minimization, Report CU-CS-240-82, Comp. Sci. Dept., University of Colorado at Boulder, 1982.

IMPORTANT: The "_f77" suffixes indicate that these routines use FORTRAN style indexing. For example, you will see

   for (i = 1; i <= n; i++)

rather than

   for (i = 0; i < n; i++)

To use the "_f77" routines you will have to declare your vectors and matrices to be one element larger (e.g., v[101] rather than v[100], and a[101][101] rather than a[100][100]), and you will have to fill elements 1 through n rather than elements 0 through n - 1. Versions of these programs that use C/Java style indexing will eventually be available. They will end with the suffix "_j".

This class was translated by a statistician from a FORTRAN version of UNCMIN. It is NOT an official translation. It wastes memory by failing to use the first elements of vectors. When public domain Java optimization routines become available from the people who produced UNCMIN, then THE CODE PRODUCED BY THE NUMERICAL ANALYSTS SHOULD BE USED.

Meanwhile, if you have suggestions for improving this code, please contact Steve Verrill at steve@ws13.fpl.fs.fed.us.

Version:

.5 --- April 14, 1998 Code was altered to Java array format. Uncmin now has a FunctionController to assemble the parallel evaluate, gradient, and hessian calls., 0.71 --- May 6, 2010

Author:

(translator)Steve Verrill, Andreas Maier

Constructor Summary

UncminForJava(FunctionController controller)

Method Summary

static void	initialize(int n, double[] x_in_java, double[] typsiz, double[] fscale, int[] method, int[] iexp, int[] msg, int[] ndigit, int[] itnlim, int[] iagflg, int[] iahflg, double[] dlt, double[] gradtl, double[] stepmx, double[] steptl) Deprecated.
void	optimizeFunction(int n, double[] x, OptimizableFunction minclass, double[] typsiz, double[] fscale, int[] method, int[] iexp, int[] msg, int[] ndigit, int[] itnlim, int[] iagflg, int[] iahflg, double[] dlt, double[] gradtl, double[] stepmx, double[] steptl, double[] xpls, double[] fpls, double[] gpls, int[] itrmcd, double[][] a, double[] udiag, double[] numericalGradient, double[] p, double[] sx, double[] wrk0, double[] wrk1, double[] wrk2, double[] wrk3) Deprecated.
void	optimizeFunction0(int dimension, double[] initialX_in_java, OptimizableFunction function, double[] vectorX, double[] functionValueAtX, double[] gradientAtX, int[] terminationCode, double[][] hessianAtX, double[] diagonalOfHessian) Deprecated.
void	optimizeFunction7(int n, double[] x_in_java, OptimizableFunction minclass, double[] typsiz, double[] fscale, int[] method, int[] iexp, int[] msg, int[] ndigit, int[] itnlim, int[] iagflg, int[] iahflg, double[] dlt, double[] gradtl, double[] stepmx, double[] steptl, double[] xpls, double[] fpls, double[] gpls, int[] itrmcd, double[][] a, double[] udiag) Deprecated.

Methods inherited from class java.lang.Object

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

Constructor Detail

UncminForJava

public UncminForJava(FunctionController controller)

Method Detail

optimizeFunction0

@Deprecated
public void optimizeFunction0(int dimension,
                                         double[] initialX_in_java,
                                         OptimizableFunction function,
                                         double[] vectorX,
                                         double[] functionValueAtX,
                                         double[] gradientAtX,
                                         int[] terminationCode,
                                         double[][] hessianAtX,
                                         double[] diagonalOfHessian)

Deprecated.

The optif0_f77 method minimizes a smooth nonlinear function of n variables. A method that computes the function value at any point must be supplied. (See Uncmin_methods.java and UncminTest.java.) Derivative values are not required. The optif0_f77 method provides the simplest user access to the UNCMIN minimization routines. Without a recompile, the user has no control over options. For details, see the Schnabel et al reference and the comments in the code. Translated by Steve Verrill, August 4, 1998.

Parameters:

dimension - The number of arguments of the function to minimize

initialX - The initial estimate of the minimum point

function - A class that implements the OptimizableFunction interface (see the definition in OptimizableFunction.java). See UncminTest_f77.java for an example of such a class. The class must define: 1.) a method, evaluate, to minimize. evaluate must have the form public static double evaluate(double x[]) where x is the vector of arguments to the function and the return value is the value of the function evaluated at x. 2.) a method, gradient, that has the form public static double [] gradient(double x[]) where the return value is the gradient of f evaluated at x. This method will have an empty body if the user does not wish to provide an analytic estimate of the gradient. 3.) a method, hessian, that has the form public static double [] [] hessian(double x[]) where the return value is the Hessian of f evaluated at x. This method will have an empty body if the user does not wish to provide an analytic estimate of the Hessian. If the user wants Uncmin to check the Hessian, then the hessian method should only fill the lower triangle (and diagonal) of the Hessian.

vectorX - The final estimate of the minimum point

functionValueAtX - The value of f_to_minimize at xpls

gradientAtX - The gradient at the local minimum xpls

terminationCode - Termination code ITRMCD = 0: Optimal solution found ITRMCD = 1: Terminated with gradient small, xpls is probably optimal ITRMCD = 2: Terminated with stepsize small, xpls is probably optimal ITRMCD = 3: Lower point cannot be found, xpls is probably optimal ITRMCD = 4: Iteration limit (150) exceeded ITRMCD = 5: Too many large steps, function may be unbounded

hessianAtX - Workspace for the Hessian (or its estimate) and its Cholesky decomposition

diagonalOfHessian - Workspace for the diagonal of the Hessian

optimizeFunction7

@Deprecated
public void optimizeFunction7(int n,
                                         double[] x_in_java,
                                         OptimizableFunction minclass,
                                         double[] typsiz,
                                         double[] fscale,
                                         int[] method,
                                         int[] iexp,
                                         int[] msg,
                                         int[] ndigit,
                                         int[] itnlim,
                                         int[] iagflg,
                                         int[] iahflg,
                                         double[] dlt,
                                         double[] gradtl,
                                         double[] stepmx,
                                         double[] steptl,
                                         double[] xpls,
                                         double[] fpls,
                                         double[] gpls,
                                         int[] itrmcd,
                                         double[][] a,
                                         double[] udiag)

Deprecated.

The optif9_f77 method minimizes a smooth nonlinear function of n variables. A method that computes the function value at any point must be supplied. (See Uncmin_methods.java and UncminTest.java.) Derivative values are not required. The optif9 method provides complete user access to the UNCMIN minimization routines. The user has full control over options. For details, see the Schnabel et al reference and the comments in the code. Translated by Steve Verrill, August 4, 1998.

Parameters:

n - The number of arguments of the function to minimize

x - The initial estimate of the minimum point

minclass - A class that implements the OptimizableFunction interface (see the definition in GradientOptimizableFunction.java). See UncminTest_f77.java for an example of such a class. The class must define: 1.) a method, evaluate, to minimize. evaluate must have the form public static double evaluate(double x[]) where x is the vector of arguments to the function and the return value is the value of the function evaluated at x. 2.) a method, gradient, that has the form public static double [] gradient(double x[]) where the return value is the gradient of f evaluated at x. This method will have an empty body if the user does not wish to provide an analytic estimate of the gradient. 3.) a method, hessian, that has the form public static double [] [] hessian(double x[]) where the return value is the Hessian of f evaluated at x. This method will have an empty body if the user does not wish to provide an analytic estimate of the Hessian. If the user wants Uncmin to check the Hessian, then the hessian method should only fill the lower triangle (and diagonal) of the Hessian.

typsiz - Typical size for each component of x

fscale - Estimate of the scale of the objective function

method - Algorithm to use to solve the minimization problem = 1 line search = 2 double dogleg = 3 More-Hebdon

iexp - = 1 if the optimization function f_to_minimize is expensive to evaluate, = 0 otherwise. If iexp = 1, then the Hessian will be evaluated by secant update rather than analytically or by finite differences.

msg - Message to inhibit certain automatic checks and output

ndigit - Number of good digits in the minimization function

itnlim - Maximum number of allowable iterations

iagflg - = 0 if an analytic gradient is not supplied

iahflg - = 0 if an analytic Hessian is not supplied

dlt - Trust region radius

gradtl - Tolerance at which the gradient is considered close enough to zero to terminate the algorithm

stepmx - Maximum allowable step size

steptl - Relative step size at which successive iterates are considered close enough to terminate the algorithm

xpls - The final estimate of the minimum point

fpls - The value of f_to_minimize at xpls

gpls - The gradient at the local minimum xpls

itrmcd - Termination code ITRMCD = 0: Optimal solution found ITRMCD = 1: Terminated with gradient small, X is probably optimal ITRMCD = 2: Terminated with stepsize small, X is probably optimal ITRMCD = 3: Lower point cannot be found, X is probably optimal ITRMCD = 4: Iteration limit (150) exceeded ITRMCD = 5: Too many large steps, function may be unbounded

a - Workspace for the Hessian (or its estimate) and its Cholesky decomposition

udiag - Workspace for the diagonal of the Hessian

initialize

@Deprecated
public static void initialize(int n,
                                         double[] x_in_java,
                                         double[] typsiz,
                                         double[] fscale,
                                         int[] method,
                                         int[] iexp,
                                         int[] msg,
                                         int[] ndigit,
                                         int[] itnlim,
                                         int[] iagflg,
                                         int[] iahflg,
                                         double[] dlt,
                                         double[] gradtl,
                                         double[] stepmx,
                                         double[] steptl)

Deprecated.

The dfault_f77 method sets default values for each input variable to the minimization algorithm. Translated by Steve Verrill, August 4, 1998.

Parameters:

n - Dimension of the problem

x - Initial estimate of the solution (to compute max step size)

typsiz - Typical size for each component of x

fscale - Estimate of the scale of the minimization function

method - Algorithm to use to solve the minimization problem

iexp - = 0 if the minimization function is not expensive to evaluate

msg - Message to inhibit certain automatic checks and output

ndigit - Number of good digits in the minimization function

itnlim - Maximum number of allowable iterations

iagflg - = 0 if an analytic gradient is not supplied

iahflg - = 0 if an analytic Hessian is not supplied

dlt - Trust region radius

gradtl - Tolerance at which the gradient is considered close enough to zero to terminate the algorithm

stepmx - "Value of zero to trip default maximum in optchk"

steptl - Tolerance at which successive iterates are considered close enough to terminate the algorithm

optimizeFunction

@Deprecated
public void optimizeFunction(int n,
                                        double[] x,
                                        OptimizableFunction minclass,
                                        double[] typsiz,
                                        double[] fscale,
                                        int[] method,
                                        int[] iexp,
                                        int[] msg,
                                        int[] ndigit,
                                        int[] itnlim,
                                        int[] iagflg,
                                        int[] iahflg,
                                        double[] dlt,
                                        double[] gradtl,
                                        double[] stepmx,
                                        double[] steptl,
                                        double[] xpls,
                                        double[] fpls,
                                        double[] gpls,
                                        int[] itrmcd,
                                        double[][] a,
                                        double[] udiag,
                                        double[] numericalGradient,
                                        double[] p,
                                        double[] sx,
                                        double[] wrk0,
                                        double[] wrk1,
                                        double[] wrk2,
                                        double[] wrk3)

Deprecated.

The optdrv_f77 method is the driver for the nonlinear optimization problem. Translated by Steve Verrill, May 18, 1998.

Parameters:

n - The dimension of the problem

x - On entry, estimate of the location of a minimum of f_to_minimize

minclass - A class that implements the OptimizableFunction interface (see the definition in GradientOptimizableFunction.java). See UncminTest_f77.java for an example of such a class. The class must define: 1.) a method, evaluate, to minimize. evaluate must have the form public static double evaluate(double x[]) where x is the vector of arguments to the function and the return value is the value of the function evaluated at x. 2.) a method, gradient, that has the form public static double [] gradient(double x[]) where the return value is the gradient of f evaluated at x. This method will have an empty body if the user does not wish to provide an analytic estimate of the gradient. 3.) a method, hessian, that has the form public static double [] [] hessian(double x[]) where the return value is the Hessian of f evaluated at x. This method will have an empty body if the user does not wish to provide an analytic estimate of the Hessian. If the user wants Uncmin to check the Hessian, then the hessian method should only fill the lower triangle (and diagonal) of the Hessian.

typsiz - Typical size of each component of x

fscale - Estimate of scale of objective function

method - Algorithm indicator 1 -- line search 2 -- double dogleg 3 -- More-Hebdon

iexp - Expense flag. 1 -- optimization function, f_to_minimize, is expensive to evaluate 0 -- otherwise If iexp = 1, the Hessian will be evaluated by secant update rather than analytically or by finite differences.

msg - On input: (> 0) message to inhibit certain automatic checks On output: (< 0) error code (= 0, no error)

ndigit - Number of good digits in the optimization function

itnlim - Maximum number of allowable iterations

iagflg - = 1 if an analytic gradient is supplied

iahflg - = 1 if an analytic Hessian is supplied

dlt - Trust region radius

gradtl - Tolerance at which the gradient is considered close enough to zero to terminate the algorithm

stepmx - Maximum step size

steptl - Relative step size at which successive iterates are considered close enough to terminate the algorithm

xpls - On exit: xpls is a local minimum

fpls - On exit: function value at xpls

gpls - On exit: gradient at xpls

itrmcd - Termination code

a - workspace for Hessian (or its approximation) and its Cholesky decomposition

udiag - workspace (for diagonal of Hessian)

numericalGradient - workspace (for gradient at current iterate)

p - workspace for step

sx - workspace (for scaling vector)

wrk0 - workspace

wrk1 - workspace

wrk2 - workspace

wrk3 - workspace