iterator
Class ISIS

java.lang.Object
  extended by iterator.AbstractIterator
      extended by iterator.AbstractOptimizer<T>
          extended by iterator.AbstractInitialSampleBasedOptimizer<T>
              extended by iterator.AbstractPopulationOptimizer<BoundConstrainedProblem>
                  extended by iterator.ISIS
All Implemented Interfaces:
DataListener, InitialSamplesSettable, Iterable, LoggableIterator, RepeatableLoggableIterator

public class ISIS
extends AbstractPopulationOptimizer<BoundConstrainedProblem>

Prof Ilan Kroo's Incredibly Simple Iterative Sampler. Can be thought of as a trust-region method performed using any monotonic fit, and a uniform sampling technique instead of a design stencil.

Author:
dgorur

Field Summary
 
Fields inherited from class iterator.AbstractPopulationOptimizer
currentBestG, lastBestG, nextBlock
 
Fields inherited from class iterator.AbstractInitialSampleBasedOptimizer
uniform
 
Fields inherited from class iterator.AbstractOptimizer
doubleFormat, funEvalCount, logWriter
 
Fields inherited from class iterator.AbstractIterator
diagnosticString, iter
 
Constructor Summary
ISIS()
           
 
Method Summary
protected  cern.colt.matrix.DoubleMatrix2D computeNextPoints()
          Computes the next set of candidate solutions.
 double getExpandFactor()
          Returns the expand factor: the factor by which to grow the trust region in case of improvement.
 double getShrinkFactor()
          Returns the shrink factor: the factor by which to grow the trust region in case of no improvement.
 void init()
          Initialization routine
static void main(java.lang.String[] args)
          Runs a test problem.
 void setExpandFactor(double expandFactor)
          Sets the expand factor.
 void setShrinkFactor(double shrinkFactor)
          Sets the shrink factor.
 
Methods inherited from class iterator.AbstractPopulationOptimizer
evaluatePoints, getPopulationSize, setInitialSamples, setInitialSamples, setInitialSamples, setPopulationSize, singleIteration
 
Methods inherited from class iterator.AbstractInitialSampleBasedOptimizer
getRandomSeed, makeInitialSamples, setRandomSeed
 
Methods inherited from class iterator.AbstractOptimizer
clear, dataAdded, dataAdded, dataRemoved, dataRemoved, getDataHandler, getFunEvalCount, getLogFileName, getMaxFunEval, getProblem, isTerminated, iterate, setDataHandler, setFunEvalCount, setLogFileName, setMaxFunEval, setProblem, updateDiagnostics, writeInitialRunLog
 
Methods inherited from class iterator.AbstractIterator
isVerbose, printDiagnostics, setVerbose
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 
Methods inherited from interface iterator.LoggableIterator
getLogFileName, setLogFileName
 
Methods inherited from interface iterator.Iterable
isTerminated, iterate
 

Constructor Detail

ISIS

public ISIS()
Method Detail

main

public static void main(java.lang.String[] args)
Runs a test problem.

Parameters:
args - None used.

getExpandFactor

public double getExpandFactor()
Returns the expand factor: the factor by which to grow the trust region in case of improvement.

Returns:
the expand factor.

getShrinkFactor

public double getShrinkFactor()
Returns the shrink factor: the factor by which to grow the trust region in case of no improvement. Note: this is the factor by which to grow the trust region, and therefore is a number smaller than unity.

Returns:
the shrink factor.

init

public void init()
Description copied from interface: Iterable
Initialization routine

Specified by:
init in interface Iterable
Overrides:
init in class AbstractPopulationOptimizer<BoundConstrainedProblem>

setExpandFactor

public void setExpandFactor(double expandFactor)
Sets the expand factor.

Parameters:
expandFactor - the given expand factor.

setShrinkFactor

public void setShrinkFactor(double shrinkFactor)
Sets the shrink factor.

Parameters:
shrinkFactor - the given shrink factor.

computeNextPoints

protected cern.colt.matrix.DoubleMatrix2D computeNextPoints()
Description copied from class: AbstractPopulationOptimizer
Computes the next set of candidate solutions.

Specified by:
computeNextPoints in class AbstractPopulationOptimizer<BoundConstrainedProblem>
Returns:
a DoubleMatrix2D containing design points row-wise.