Package xal.extension.solver.algorithm

Class RandomShrinkSearch

java.lang.Object

xal.extension.solver.algorithm.SearchAlgorithm

xal.extension.solver.algorithm.RandomShrinkSearch

All Implemented Interfaces:

AlgorithmScheduleListener, SolutionJudgeListener

public class RandomShrinkSearch
extends SearchAlgorithm

RandomSearchAlgorithm looks for points bounded by the specified variable limits. Every iteration a subset of variables are randomly chosen to be changed. The value chosen for each variable is randomly selected from a uniform distribution within a window for that variable. The window begins as the full limit range allowed by the variable. Whenever a better solution is found, the window changes so that it is three times wider than the distance between the last best point and the new best point. The window gets centered around the new best points. Then the window gets clipped as necessary to satisfy the variable limits. Note that the window may grow or shrink depending on the distance between successive top points. As the optimal solution is approached, the window should tend to shrink around the optimal point. Note that each variable has its own window and the window automatically adjusts to the variable according to the progress.

Author:

t6p

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
class	RandomShrinkSearch.ComboSearcher	Use a combination of search engines to search for the best solution.
protected class	RandomShrinkSearch.RandomSearcher	A searcher that performs a simple random search in the entire search space.
protected static interface	RandomShrinkSearch.Searcher	Interface for classes that search for solutions.
protected class	RandomShrinkSearch.ShrinkSearcher	ShrinkSearcher searches for the next trial point by adjusting the search domain per variable depending on how much a variable has changed between the best solutions found so far.

Field Summary

Fields

Modifier and Type	Field	Description
protected TrialPoint	bestPoint	The current best point.
protected double	bestSatisfaction	keeps track of internal best Satisfaction
protected boolean	isLastEvaluation	keeps track if the last run is being executed
protected RandomShrinkSearch.Searcher	searcher	The active search technique (random or shrink).

Fields inherited from class xal.extension.solver.algorithm.SearchAlgorithm

EFFICIENCY_COMPARATOR, problem

Constructor Summary

Constructors

Constructor	Description
RandomShrinkSearch()

Method Summary

Modifier and Type	Method	Description
void	algorithmAvailable(SearchAlgorithm source)	Handle a message that an algorithm is available.
void	algorithmUnavailable(SearchAlgorithm source)	Handle a message that an algorithm is not available.
void	foundNewOptimalSolution(SolutionJudge source, List<Trial> solutions, Trial solution)	Handle a message that a new optimal solution has been found.
String	getLabel()	Get the label for this search algorithm.
int	globalRating()	Get the rating for this algorithm which in an integer between 0 and 10 and indicates how well this algorithm performs on global searches.
int	localRating()	Get the rating for this algorithm which in an integer between 0 and 10 and indicates how well this algorithm performs on local searches.
TrialPoint	nextTrialPoint()	Calculate and get the next trial point to evaluate.
void	performRun(AlgorithmSchedule algorithmSchedule)	Calculate the next few trial points.
void	reset()	reset for searching from scratch; forget history
protected void	resetBestPoint()	Reset the trial point's variables to their starting values.
void	trialScored(AlgorithmSchedule schedule, Trial trial)	Handle a message that a trial has been scored.

Methods inherited from class xal.extension.solver.algorithm.SearchAlgorithm

addSearchAlgorithmListener, algorithmRunExecuted, algorithmRunWillExecute, evaluateTrialPoint, executeRun, getEfficiency, getEvaluationsLeft, getMaxEvaluationsPerRun, getMinEvaluationsPerRun, removeSearchAlgorithmListener, setProblem, setProposedEvaluations, trialVetoed

Methods inherited from class java.lang.Object

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

Field Details

bestPoint

protected TrialPoint bestPoint

The current best point.

searcher

protected RandomShrinkSearch.Searcher searcher

The active search technique (random or shrink).

bestSatisfaction

protected double bestSatisfaction

keeps track of internal best Satisfaction

isLastEvaluation

protected boolean isLastEvaluation

keeps track if the last run is being executed

Constructor Details

RandomShrinkSearch

public RandomShrinkSearch()

Method Details

getLabel

public String getLabel()

Get the label for this search algorithm.

Specified by:

getLabel in class SearchAlgorithm

Returns:

The label for this algorithm

reset

public void reset()

reset for searching from scratch; forget history

Overrides:

reset in class SearchAlgorithm

resetBestPoint

protected void resetBestPoint()

Reset the trial point's variables to their starting values.

performRun

public void performRun(AlgorithmSchedule algorithmSchedule)

Calculate the next few trial points.

Specified by:

performRun in class SearchAlgorithm

Parameters:

algorithmSchedule - the algorithm run to perform the evaluation

nextTrialPoint

public TrialPoint nextTrialPoint()

Calculate and get the next trial point to evaluate.

Returns:

The next trial point to evaluate.

globalRating

public int globalRating()

Get the rating for this algorithm which in an integer between 0 and 10 and indicates how well this algorithm performs on global searches.

Returns:

The global search rating for this algorithm.

localRating

public int localRating()

Get the rating for this algorithm which in an integer between 0 and 10 and indicates how well this algorithm performs on local searches.

Returns:

The local search rating for this algorithm.

algorithmAvailable

public void algorithmAvailable(SearchAlgorithm source)

Handle a message that an algorithm is available.

Parameters:

source - The source of the available algorithm.

algorithmUnavailable

public void algorithmUnavailable(SearchAlgorithm source)

Handle a message that an algorithm is not available.

Parameters:

source - The source of the available algorithm.

trialScored

public void trialScored(AlgorithmSchedule schedule,
 Trial trial)

Handle a message that a trial has been scored.

Specified by:

trialScored in interface AlgorithmScheduleListener

Overrides:

trialScored in class SearchAlgorithm

Parameters:

schedule - Description of the Parameter

trial - Description of the Parameter This is for operating this algorithm completely independently of any other algorithm

foundNewOptimalSolution

public void foundNewOptimalSolution(SolutionJudge source,
 List<Trial> solutions,
 Trial solution)

Handle a message that a new optimal solution has been found.

Specified by:

foundNewOptimalSolution in interface SolutionJudgeListener

Overrides:

foundNewOptimalSolution in class SearchAlgorithm

Parameters:

source - The source of the new optimal solution.

solutions - The list of solutions.

solution - The new optimal solution.