com.dna.jopt.util.costconverger

Interface ICostConverger

All Known Implementing Classes:

AbstractCostConverger, JoinedCostConverger, JoinedCostConvergerWithFinalAutoFilterStep

public interface ICostConverger

This interface of the converger allows to set an iteration treshold after which the Optimisation is stopped if the cost does not change. The converger can be customized to stop the Optimisation if the distance or time do not change even tough the cost is still fluctuating.

Implementation example:

 // Create external CostConverger
 ICostConverger myJoinedCostConverger = new JoinedCostConverger(this);
 myJoinedCostConverger.addExecutionAlgorithm(GeneticEvolutionAlgo.MY_TITLE);
 myJoinedCostConverger.setConvergenceThreshold(5);

 CompletableFuture<IOptimizationResult> resultFuture = this.startRunAsync();

 // It is important to block the call, otherwise optimization will be terminated
 try {
   resultFuture.get();
 } catch (InterruptedException | ExecutionException e) {
   e.printStackTrace();
 }
 

Since:

03/08/2019

Version:

13/09/2019

Author:

DNA

Method Summary

Modifier and Type	Method and Description
void	addExecutionAlgorithm(String exectutionAlgo) Defines where the Optimization can be stopped by the Converger.
double	getConvergenceCost(IOptimizationProgress curP)
CompletableFuture<Void>	onConverged(IOptimization opti)
void	onDone(IOptimization opti)
void	setConvergenceThreshold(int numProgressSteps) Sets the convergence threshold X.
void	setOnConvergedTimeOut(long onConvergedtimeout, TimeUnit onConvergedTimeUnit)

Method Detail

setConvergenceThreshold

void setConvergenceThreshold(int numProgressSteps)

Sets the convergence threshold X. If the convergence cost did not change for X iterations in % the converger stops the optimizations.

Parameters:

numProgressSteps - the new convergence threshold

addExecutionAlgorithm

void addExecutionAlgorithm(String exectutionAlgo)

Defines where the Optimization can be stopped by the Converger. For implementation example please see ICostConverger.

Possible values are "GeneticEvolution" and "SimulatedAnnealing".

Parameters:

exectutionAlgo - the string of the algorithm in which the optimization can be stopped safely

getConvergenceCost

double getConvergenceCost(IOptimizationProgress curP)

onDone

void onDone(IOptimization opti)

onConverged

CompletableFuture<Void> onConverged(IOptimization opti)

setOnConvergedTimeOut

void setOnConvergedTimeOut(long onConvergedtimeout,
                           TimeUnit onConvergedTimeUnit)