edu.stanford.nlp.optimization
Class CGMinimizer

java.lang.Object
  |
  +--edu.stanford.nlp.optimization.CGMinimizer

All Implemented Interfaces:

Minimizer

public class CGMinimizer

extends Object

implements Minimizer

Conjugate-gradient implementation based on the code in Numerical Recipes in C. (See p. 423 and others.) As of now, it requires a differentiable function (DiffFunction) as input. Equality constraints are supported; inequality constraints may soon be added. The basic way to use the minimizer is with a null constructor, then the simple minimize method:

Minimizer cgm = new CGMinimizer();
DiffFunction df = new SomeDiffFunction();
double tol = 1e-4;
double[] initial = getInitialGuess();
double[] minimum = cgm.minimize(df,tol,initial);

Since:

1.0

Constructor Summary

CGMinimizer()
Basic constructor, use this.

CGMinimizer(boolean silent)
Pass in false to get per-iteration progress reports (to stderr).

CGMinimizer(Function monitor)
Perform minimization with monitoring.

Method Summary

double[]

minimize(Function function, double functionTolerance, double[] initial) Attempts to find an unconstrained minimum of the objective function starting at initial, within functionTolerance.

Methods inherited from class java.lang.Object

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

Constructor Detail

CGMinimizer

public CGMinimizer()

Basic constructor, use this.

CGMinimizer

public CGMinimizer(boolean silent)

Pass in false to get per-iteration progress reports (to stderr).

Parameters:

silent - a boolean value

CGMinimizer

public CGMinimizer(Function monitor)

Perform minimization with monitoring. After each iteration, monitor.valueAt(x) gets called, with the double array x being that iteration's ending point. A return < tol forces convergence (terminates the CG procedure). Specially for Kristina.

Parameters:

monitor - a Function value

Method Detail

minimize

public double[] minimize(Function function,
                         double functionTolerance,
                         double[] initial)

Description copied from interface: Minimizer

Attempts to find an unconstrained minimum of the objective function starting at initial, within functionTolerance.

Specified by:

minimize in interface Minimizer

Parameters:

function - the objective function

functionTolerance - a double value

initial - a initial feasible point