edu.jhu.bme.smile.commons.optimize
Class LevenbergMarquardt

java.lang.Object
  edu.jhu.bme.smile.commons.optimize.LevenbergMarquardt

All Implemented Interfaces:

OptimizerNDContinuousDifferentiable

public class LevenbergMarquardt
extends java.lang.Object
implements OptimizerNDContinuousDifferentiable

This optimizer uses the Levenberg Marquardt method in order to find the minimum of a function. You can choose an initial starting point to start the optimization from. Code adapted from http://www.mathworks.com/matlabcentral/fx_files/22778/1/content/optimization/seeMe.htm

Author:

Hanlin Wan, Yufeng Guo

Constructor Summary

LevenbergMarquardt()
Default constructor for LM.

LevenbergMarquardt(int maxIter)
Constructor to let you set the maximum number of iterations.

Method Summary

double[]	getExtrema() Gets the extrema that was found.
int	getIterations() Gets the number of iterations done.
java.util.ArrayList<double[][]>	getLog() Gets the logged data
void	initialize(OptimizableNDContinuousDifferentiable function) Initializes the optimizer with a differentiable ND function.
void	initialize(OptimizableNDContinuousDifferentiable function, double[] init) Initializes the optimizer with a differentiable ND function.
void	logData(boolean turnOn, int maxSteps) Turns on logging for each iteration of the optimization.
boolean	optimize(boolean findMinima) Optimizes the function.
void	setGamma(double lamInit, double lamMax) Sets the initial and maximum lambda values.
java.lang.String	statusMessage() Gets the status message.

Methods inherited from class java.lang.Object

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

Constructor Detail

LevenbergMarquardt

public LevenbergMarquardt()

Default constructor for LM. Maximum iterations set to 1000.

LevenbergMarquardt

public LevenbergMarquardt(int maxIter)

Constructor to let you set the maximum number of iterations.

Parameters:

maxIter - maximum number of iterations

Method Detail

logData

public void logData(boolean turnOn,
                    int maxSteps)

Turns on logging for each iteration of the optimization.

Parameters:

turnOn - true if logging, false if not

maxSteps - maximum number of steps to log

getLog

public java.util.ArrayList<double[][]> getLog()

Gets the logged data

Returns:

array of logged data

getExtrema

public double[] getExtrema()

Gets the extrema that was found.

Specified by:

getExtrema in interface OptimizerNDContinuousDifferentiable

Returns:

the extrema array of where the extrema is

getIterations

public int getIterations()

Gets the number of iterations done.

Specified by:

getIterations in interface OptimizerNDContinuousDifferentiable

Returns:

number of iterations

setGamma

public void setGamma(double lamInit,
                     double lamMax)

Sets the initial and maximum lambda values. Default values are: initial lambda=0.001; maximum lambda=1000.

Parameters:

lamInit - initial lambda value

lamMax - maximum lambda value

initialize

public void initialize(OptimizableNDContinuousDifferentiable function,
                       double[] init)

Initializes the optimizer with a differentiable ND function. You can specify the starting point.

Parameters:

function - function to minimize

init - starting point

initialize

public void initialize(OptimizableNDContinuousDifferentiable function)

Initializes the optimizer with a differentiable ND function. The starting point is set to halfway between the minimum and maximum domain points in each direction.

Specified by:

initialize in interface OptimizerNDContinuousDifferentiable

Parameters:

function - function to optimize

optimize

public boolean optimize(boolean findMinima)

Optimizes the function.

Specified by:

optimize in interface OptimizerNDContinuousDifferentiable

Parameters:

findMinima - true if minimizing, false if maximizing

Returns:

whether the function was successfully optimized

statusMessage

public java.lang.String statusMessage()

Gets the status message.

Specified by:

statusMessage in interface OptimizerNDContinuousDifferentiable

Returns:

status message