halfdan/eva2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Commit of MK branch revs. 393 and 396. Updates to stats, clustering plus all new CBNEA.

Browse Source
This commit is contained in:
Marcel Kronfeld
2009-11-20 09:55:56 +00:00
parent 22c8c0e998
commit c26b394763
23 changed files with 1462 additions and 501 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/eva2/gui/BeanInspector.java
View File

@@ -164,7 +164,7 @@ public class BeanInspector {
}
if (Target instanceof List && !(Target instanceof Population)) { // handle the list case
StringBuffer sbuf = new StringBuffer("[ ");
StringBuffer sbuf = new StringBuffer("[ ");
List<?> lst = (List<?>)Target;
for (Object o : lst) {
sbuf.append(o.toString());

18
src/eva2/gui/Plot.java
View File

@@ -39,14 +39,14 @@ import javax.swing.JFrame;
import javax.swing.JOptionPane;
import javax.swing.JPanel;
import com.sun.image.codec.jpeg.JPEGCodec;
import com.sun.image.codec.jpeg.JPEGImageEncoder;
import eva2.EvAInfo;
import eva2.server.go.individuals.AbstractEAIndividual;
import eva2.server.go.populations.Population;
import eva2.tools.chart2d.DPointSet;
import eva2.tools.tool.BasicResourceLoader;
import eva2.server.go.populations.Population;
/*==========================================================================*
* CLASS DECLARATION
*==========================================================================*/
@@ -287,10 +287,22 @@ public class Plot implements PlotInterface, Serializable {
*/
public void drawPopulation(String prefix, Population pop) {
for (int i=0; i<pop.size(); i++) {
getFunctionArea().drawIcon(1, prefix+" "+pop.getEAIndividual(i).getFitness(0), pop.getEAIndividual(i).getDoublePosition(), 2);
drawIndividual(1, 2, prefix, pop.getEAIndividual(i));
}
}
/**
* Draw an individual to the Plot instance. It is annotated with the
* given prefix and its fitness.
*
* @param prefix
* @param pop
* @see FunctionArea.drawIcon
*/
public void drawIndividual(int iconType, int graphID, String prefix, AbstractEAIndividual indy) {
getFunctionArea().drawIcon(iconType, prefix+" "+indy.getFitness(0), indy.getDoublePosition(), graphID);
}
public void setPreferredSize(Dimension prefSize) {
if (m_Frame != null) {
m_Frame.setPreferredSize(prefSize);

80
src/eva2/server/go/operators/cluster/ClusteringDensityBased.java
View File

@@ -6,6 +6,7 @@ import eva2.server.go.individuals.AbstractEAIndividual;
import eva2.server.go.operators.distancemetric.InterfaceDistanceMetric;
import eva2.server.go.operators.distancemetric.PhenotypeMetric;
import eva2.server.go.populations.Population;
import eva2.tools.Pair;
/** The DBSCAN method. As far as I recall this is an hierachical
@@ -76,14 +77,7 @@ public class ClusteringDensityBased implements InterfaceClustering, java.io.Seri
return (Object) new ClusteringDensityBased(this);
}
/** This method allows you to search for clusters in a given population. The method
* returns Number of populations. The first population contains all individuals that
* could not be associated with any cluster and may be empty.
* All other populations group individuals into clusters.
* @param pop The population of individuals that is to be clustered.
* @return Population[]
*/
public Population[] cluster(Population pop) {
public Population[] cluster(Population pop, Population referencePop) {
ConnectionMatrix = new boolean[pop.size()][pop.size()];
Clustered = new boolean[pop.size()];
AbstractEAIndividual tmpIndy1, tmpIndy2;
@@ -151,36 +145,64 @@ public class ClusteringDensityBased implements InterfaceClustering, java.io.Seri
* @param species2 The second species.
* @return True if species converge, else False.
*/
public boolean mergingSpecies(Population species1, Population species2) {
public boolean mergingSpecies(Population species1, Population species2, Population referencePop) {
if (m_TestConvergingSpeciesOnBestOnly) {
if (this.m_Metric.distance(species1.getBestEAIndividual(), species2.getBestEAIndividual()) < this.m_ClusterDistance) return true;
else return false;
double specDist = this.m_Metric.distance(species1.getBestEAIndividual(), species2.getBestEAIndividual());
// System.out.println("Dist between species is " + specDist);
return (specDist < this.m_ClusterDistance);
} else {
Population tmpPop = new Population(species1.size()+species2.size());
tmpPop.addPopulation(species1);
tmpPop.addPopulation(species2);
if (this.cluster(tmpPop).length <= 2) return true;
if (this.cluster(tmpPop, referencePop).length <= 2) return true;
else return false;
}
}
/** This method decides if a unclustered individual belongs to an already established species.
* @param indy A unclustered individual.
* @param species A species.
* @return True or False.
// /** This method decides if a unclustered individual belongs to an already established species.
// * @param indy A unclustered individual.
// * @param species A species.
// * @return True or False.
// */
// public boolean belongsToSpecies(AbstractEAIndividual indy, Population species, Population pop) {
// if (this.m_TestConvergingSpeciesOnBestOnly) {
// if (this.m_Metric.distance(indy, species.getBestEAIndividual()) < this.m_ClusterDistance) return true;
// else return false;
// } else {
// Population tmpPop = (Population)species.clone();
// tmpPop.add(indy);
// if (this.cluster(tmpPop)[0].size() == 0) return true;
// else return false;
// }
// }
/**
* Try to associate a set of loners with a given set of species. Return a list
* of indices assigning loner i with species j for all loners. If no species can
* be associated, -1 is returned as individual entry.
* Note that the last cluster threshold is used which may have depended on the last
* generation.
*
* @param loners
* @param species
* @return associative list matching loners to species.
*/
public boolean belongsToSpecies(AbstractEAIndividual indy, Population species) {
if (this.m_TestConvergingSpeciesOnBestOnly) {
if (this.m_Metric.distance(indy, species.getBestEAIndividual()) < this.m_ClusterDistance) return true;
else return false;
} else {
Population tmpPop = (Population)species.clone();
tmpPop.add(indy);
if (this.cluster(tmpPop)[0].size() == 0) return true;
else return false;
}
public int[] associateLoners(Population loners, Population[] species, Population referencePop) {
int[] res = new int[loners.size()];
for (int l=0; l<loners.size(); l++) {
double minDist = -1;
res[l]=-1;
for (int spI=0; spI<species.length; spI++) { // O(species.length^2)
Pair<Integer,Double> iDist = Population.getClosestFarthestIndy(loners.getEAIndividual(l), species[spI], m_Metric, true);
if (iDist.tail() < m_ClusterDistance) { // its close enough to be added
// set SP ID only if its the closest species which is still below cluster distance
if (minDist<0 || (iDist.tail() < minDist)) res[l]=spI;
}
} // end for all species
} // end for all loners
return res;
}
/**********************************************************************************************************************
* These are for GUI
*/
@@ -236,6 +258,10 @@ public class ClusteringDensityBased implements InterfaceClustering, java.io.Seri
return "Set the minimum group size for the DBSCAN method.";
}
public String initClustering(Population pop) {
return null;
}
// /** For debuggy only
// * @param plot TopoPlot

37
src/eva2/server/go/operators/cluster/ClusteringKMeans.java
View File

@@ -1,5 +1,7 @@
package eva2.server.go.operators.cluster;
import java.util.Arrays;
import eva2.gui.Chart2DDPointIconCircle;
import eva2.gui.Chart2DDPointIconText;
import eva2.gui.GraphPointSet;
@@ -59,7 +61,7 @@ public class ClusteringKMeans implements InterfaceClustering, java.io.Serializab
* @param pop The population of individuals that is to be clustered.
* @return Population[]
*/
public Population[] cluster(Population pop) {
public Population[] cluster(Population pop, Population referencePop) {
double[][] data = this.extractClusterDataFrom(pop);
if (!(this.m_ReuseC) || (this.m_C == null)) {
this.m_C = new double[this.m_K][];
@@ -278,20 +280,27 @@ public class ClusteringKMeans implements InterfaceClustering, java.io.Serializab
* @param species2 The second species.
* @return True if species converge, else False.
*/
public boolean mergingSpecies(Population species1, Population species2) {
public boolean mergingSpecies(Population species1, Population species2, Population referencePop) {
// @todo i could use the BIC metric from X-means to calculate this
return false;
}
/** This method decides if a unclustered individual belongs to an already established species.
* @param indy A unclustered individual.
* @param species A species.
* @return True or False.
*/
public boolean belongsToSpecies(AbstractEAIndividual indy, Population species) {
// @todo perhaps the same as in convergingSpecies
return false;
}
// /** This method decides if a unclustered individual belongs to an already established species.
// * @param indy A unclustered individual.
// * @param species A species.
// * @return True or False.
// */
// public boolean belongsToSpecies(AbstractEAIndividual indy, Population species, Population pop) {
// // @todo perhaps the same as in convergingSpecies
// return false;
// }
public int[] associateLoners(Population loners, Population[] species, Population referencePop) {
int[] res=new int[loners.size()];
System.err.println("Warning, associateLoners not implemented for " + this.getClass());
Arrays.fill(res, -1);
return res;
}
/** This method allows you to recieve the c centroids
* @return The centroids
@@ -313,7 +322,7 @@ public class ClusteringKMeans implements InterfaceClustering, java.io.Serializab
f1.setProblemDimension(2);
f1.setEAIndividual(new ESIndividualDoubleData());
f1.initPopulation(pop);
ckm.cluster(pop);
ckm.cluster(pop, (Population)null);
}
@@ -374,4 +383,8 @@ public class ClusteringKMeans implements InterfaceClustering, java.io.Serializab
public String reuseCTipText() {
return "Toggel reuse of previously found cluster centroids.";
}
public String initClustering(Population pop) {
return null;
}
}

401
src/eva2/server/go/operators/cluster/ClusteringNearestBetter.java Normal file
View File

@@ -0,0 +1,401 @@
package eva2.server.go.operators.cluster;
import java.io.Serializable;
import java.util.ArrayList;
import java.util.LinkedList;
import java.util.Vector;
import eva2.gui.BeanInspector;
import eva2.gui.GenericObjectEditor;
import eva2.server.go.individuals.AbstractEAIndividual;
import eva2.server.go.individuals.AbstractEAIndividualComparator;
import eva2.server.go.operators.distancemetric.InterfaceDistanceMetric;
import eva2.server.go.operators.distancemetric.PhenotypeMetric;
import eva2.server.go.populations.Population;
import eva2.tools.Pair;
/**
* Hierarchical clustering after Preuss et al., "Counteracting Genetic Drift and Disruptive Recombination
* in (mu+,lambda)-EA on Multimodal Fitness Landscapes", GECCO '05.
*
* A tree is produced by assigning each individual the closest individual with better fitness.
* Connections with a distance above a certain threshold are cut. After that, each interconnected subtree forms a cluster.
* In the paper, the threshold is deduced as 2*d_p for d_p: the mean distance in the population.
*
* @author mkron
*
*/
public class ClusteringNearestBetter implements InterfaceClustering, Serializable {
private static final long serialVersionUID = 1L;
private InterfaceDistanceMetric metric = new PhenotypeMetric();
private double absoluteDistThreshold = 0.5;
private boolean thresholdMultipleOfMeanDist = true;
private double meanDistFactor = 2.; // recommended setting
private double currentMeanDistance = -1.;
private int minimumGroupSize = 3;
private boolean testConvergingSpeciesOnBestOnly = true; // if two species are tested for convergence, only the best indies may be compared regarding the distance threshold
private int[] uplink;
private double[] uplinkDist;
private AbstractEAIndividualComparator comparator = new AbstractEAIndividualComparator();
private Vector<Integer>[] children;
private static final String initializedForKey = "initializedClustNearestBetterOnHash";
private static final String initializedRefData = "initializedClustNearestBetterData";
private static boolean TRACE = true;
public ClusteringNearestBetter() {
}
public ClusteringNearestBetter(ClusteringNearestBetter o) {
this.metric = o.metric;
this.absoluteDistThreshold = o.absoluteDistThreshold;
this.thresholdMultipleOfMeanDist = o.thresholdMultipleOfMeanDist;
this.meanDistFactor = o.meanDistFactor;
this.currentMeanDistance = o.currentMeanDistance;
this.minimumGroupSize = o.minimumGroupSize;
this.comparator = (AbstractEAIndividualComparator)o.comparator.clone();
this.testConvergingSpeciesOnBestOnly = o.testConvergingSpeciesOnBestOnly;
}
public void hideHideable() {
setAdaptiveThreshold(isAdaptiveThreshold());
}
/** This method allows you to make a deep clone of
* the object
* @return the deep clone
*/
public Object clone() {
return (Object) new ClusteringNearestBetter(this);
}
/**
* Try to associate a set of loners with a given set of species. Return a list
* of indices assigning loner i with species j for all loners. If no species can
* be associated, -1 is returned as individual entry.
* Note that the last cluster threshold is used which may have depended on the last
* generation.
*
* @param loners
* @param species
* @return associative list matching loners to species.
*/
public int[] associateLoners(Population loners, Population[] species, Population referenceSet) {
// Pair<Integer,Double>[][] closestPerSpecList = new Pair[loners.size()][species.length];
int[] res = new int[loners.size()];
getRefData(referenceSet, loners);
for (int l=0; l<loners.size(); l++) { // for each loner: search closest better indy for each species.
int nearestBetterSpeciesID=-1;
double nearestBetterDist=-1;
for (int spI=0; spI<species.length; spI++) { // loop species
boolean lonerIndyIsBest = (comparator.compare(loners.getEAIndividual(l), species[spI].getBestEAIndividual())<=0);
if (lonerIndyIsBest) { // if the loner is the best, check the distance to the best indy within the species
double curDist = metric.distance(loners.getEAIndividual(l), species[spI].getBestEAIndividual());
//Population.getClosestFarthestIndy(loners.getEAIndividual(l), species[spI], metric, true).tail();
if (nearestBetterDist<0 || (curDist < nearestBetterDist)) {
// System.out.println("Loner is better " + loners.getEAIndividual(l) + " than best " + species[spI].getBestEAIndividual() + ", dist is "+curDist);
nearestBetterSpeciesID=spI;
nearestBetterDist = curDist;
}
} else {
for (int i=0; i<species[spI].size(); i++) { //loop indies in species
double curDist = metric.distance(loners.getEAIndividual(l), species[spI].getEAIndividual(i));
boolean specIndyIsBetter = (comparator.compare(species[spI].getEAIndividual(i), loners.getEAIndividual(l))<0);
if (specIndyIsBetter && (nearestBetterDist<0 || (curDist < nearestBetterDist))) {
// if the found indy species is better than the loner, it is a possible cluster.
// store the closest possible cluster.
// nearestBetterIndyID = i;
nearestBetterSpeciesID=spI;
nearestBetterDist = curDist;
}
}
}
// if (comparator.compare(species[spI].getEAIndividual(closestID), loners.getEAIndividual(l))<0) {
//
// if (closestClustDist<0 || (closestDist < closestClustDist)) {
// closestClustDist = closestDist;
// closestClustID = spI;
// }
// }
} // end loop species
if (nearestBetterDist < currentDistThreshold()) {
// System.out.println("dist is " + nearestBetterDist + ", assigning spec " + nearestBetterSpeciesID);
res[l]=nearestBetterSpeciesID;
} else res[l]=-1;
} // end for all loners
return res;
}
// public boolean belongsToSpecies(AbstractEAIndividual indy,
// Population species, Population pop) {
// // this sucks since every time the full clustering must be performed...
// return false;
//// if (thresholdMultipleOfMeanDist) currentMeanDistance = pop.getPopulationMeasures(metric)[0];
//// ArrayList<AbstractEAIndividual> sorted = pop.getSorted(comparator);
//// for (int i=sorted.size()-1; i>=1; i--) { // start with worst indies
//// if (sorted.get(i).getIndyID()==indy.getIndyID()) { // found the desired indy.
//// int uplink=-1; double uplinkDist = -1;
//// for (int j=i-1; j>=0; j--) { // search nearest better indy
//// double curDist = metric.distance(sorted.get(i), sorted.get(j));
//// if (uplinkDist<0 || (curDist < uplinkDist)) {
//// uplink = j;
//// uplinkDist = curDist;
//// }
//// }
//// // if it belongs to species spec and the distance is below threshold, be happy and return true
//// if (uplink==-1) { // it is the best individual?
//// return false;
//// }
//// if (uplinkDist > currentDistThreshold()) return false;
//// else {
//// return (species.isMemberByID(pop.getEAIndividual(uplink)));
//// }
//// }
//// }
//// // size <= 1?
//// return false;
// }
/**
* Perform one clustering step to measure the mean distance to the
* nearest better individual (only if used).
*/
public String initClustering(Population pop) {
if (this.isAdaptiveThreshold()) {
ArrayList<AbstractEAIndividual> sorted = pop.getSorted(comparator);
if (uplink==null || (uplink.length!=pop.size())) uplink = new int[pop.size()]; // parent index of all indys
if (uplinkDist==null || (uplinkDist.length!=pop.size())) uplinkDist = new double[pop.size()]; // parent distance for all indys
if (children==null || (children.length!=pop.size())) children = new Vector[pop.size()]; // list of children for all indies
else if (children.length==pop.size()) for (int i=0; i<pop.size(); i++) children[i]=null;
currentMeanDistance = createClusterTreeFromSortedPop(sorted);
if (TRACE) pop.putData(initializedForKey, pop.hashCode());
pop.putData(initializedRefData, currentMeanDistance);
return initializedRefData;
} else return null;
}
public Population[] cluster(Population pop, Population referenceSet) {
if (pop.isEmpty()) return new Population[]{pop.cloneWithoutInds()};
ArrayList<AbstractEAIndividual> sorted = pop.getSorted(comparator);
if (uplink==null || (uplink.length!=pop.size())) uplink = new int[pop.size()]; // parent index of all indys
if (uplinkDist==null || (uplinkDist.length!=pop.size())) uplinkDist = new double[pop.size()]; // parent distance for all indys
if (children==null || (children.length!=pop.size())) children = new Vector[pop.size()]; // list of children for all indies
else if (children.length==pop.size()) for (int i=0; i<pop.size(); i++) children[i]=null;
if (TRACE) {
System.out.println("Current pop measures: " + BeanInspector.toString(pop.getPopulationMeasures(metric)[0]));
System.out.println("Current threshold: " + currentDistThreshold());
}
if (isAdaptiveThreshold()) { // test if there was a valid initialization step
if (!getRefData(referenceSet, pop)) currentMeanDistance=createClusterTreeFromSortedPop(sorted);
else createClusterTreeFromSortedPop(sorted);
} else createClusterTreeFromSortedPop(sorted);
// now go through indies starting with best.
// Add all children which are closer than threshold and recursively their children to a cluster.
// Mark them as clustered and start with the next best unclustered.
int current = 0; // top indy is first
boolean[] clustered = new boolean[pop.size()];
LinkedList<Population> allClusters = new LinkedList<Population>();
while (current<sorted.size()) {
Population currentClust = pop.cloneWithoutInds();
currentClust.add(sorted.get(current));
clustered[current]=true;
addChildren(current, clustered, sorted, currentClust);
// currentClust now recursively contains all children - the cluster is complete
// now jump to the next best unclustered indy
allClusters.add(currentClust);
while (current<sorted.size() && (clustered[current])) current++;
}
ArrayList<Population> finalClusts = new ArrayList<Population>(allClusters.size());
finalClusts.add(pop.cloneWithoutInds());
for (Population clust : allClusters) {
if (clust.size()<minimumGroupSize) { // add to loner population
finalClusts.get(0).addPopulation(clust);
} else { // add to cluster list
finalClusts.add(clust);
}
}
Population[] finalArr = new Population[finalClusts.size()];
return finalClusts.toArray(finalArr);
}
/**
* Get the reference data from a population instance that should have been initialized.
* If the reference set is null, the backup is treated as reference set.
*
* @param referenceSet
* @param backup
*/
private boolean getRefData(Population referenceSet, Population backup) {
if (referenceSet==null) referenceSet=backup;
Double refDat = (Double)referenceSet.getData(initializedRefData);
if (refDat!=null) {
if (TRACE) { // check hash
Integer hash=(Integer)referenceSet.getData(initializedForKey);
if ((hash==null) || (hash!=referenceSet.hashCode())) {
System.err.println("Warning, missing initialization before clustering for ClusteringNearestBetter!");
return false;
}
}
currentMeanDistance = refDat.doubleValue();
return true;
} else {
System.err.println("Warning, missing reference data - forgot reference set initialization? " + this.getClass());
return false;
}
}
private double createClusterTreeFromSortedPop(ArrayList<AbstractEAIndividual> sorted) {
double edgeLengthSum=0; int edgeCnt = 0;
for (int i=sorted.size()-1; i>=1; i--) { // start with worst indies
// search for closest indy which is better
uplink[i]=-1;
uplinkDist[i] = -1;
for (int j=i-1; j>=0; j--) { // look at all which are better
// if the j-th indy is closer, reset the index
double curDist = metric.distance(sorted.get(i), sorted.get(j));
if (uplinkDist[i]<0 || (curDist < uplinkDist[i])) {
uplink[i] = j;
uplinkDist[i] = curDist;
}
}
// the closest best for indy i is now known. connect them in the graph.
if (children[uplink[i]]==null) children[uplink[i]]=new Vector<Integer>();
children[uplink[i]].add(i);
edgeLengthSum+=uplinkDist[i];
edgeCnt++;
}
// currentMeanDistance = pop.getPopulationMeasures(metric)[0];
return edgeLengthSum/((double)edgeCnt); // the average edge length
}
/**
* Add the next layer of children to the clustered population.
*
* @param current
* @param clustered
* @param sorted
* @param currentClust
*/
private void addChildren(int current, boolean[] clustered, ArrayList<AbstractEAIndividual> sorted, Population currentClust) {
if (children[current]!=null && (children[current].size()>0)) {
for (int i=0; i<children[current].size(); i++) {
if ((!clustered[children[current].get(i)]) && (uplinkDist[children[current].get(i)] < currentDistThreshold())) {
// the first child is not clustered yet and below distance threshold.
// so add it to the cluster, mark it, and proceed recursively.
currentClust.add(sorted.get(children[current].get(i)));
clustered[children[current].get(i)]=true;
addChildren(children[current].get(i), clustered, sorted, currentClust);
}
}
} else {
// nothing more to do
}
}
private double currentDistThreshold() {
if (thresholdMultipleOfMeanDist) return meanDistFactor*currentMeanDistance;
else return absoluteDistThreshold;
}
/** This method allows you to decide if two species converge.
* @param species1 The first species.
* @param species2 The second species.
* @return True if species converge, else False.
*/
public boolean mergingSpecies(Population species1, Population species2, Population referenceSet) {
getRefData(referenceSet, species1);
if (testConvergingSpeciesOnBestOnly) {
if (this.metric.distance(species1.getBestEAIndividual(), species2.getBestEAIndividual()) < this.currentDistThreshold()) return true;
else return false;
} else {
Population tmpPop = new Population(species1.size()+species2.size());
tmpPop.addPopulation(species1);
tmpPop.addPopulation(species2);
if (this.cluster(tmpPop, referenceSet).length <= 2) return true;
else return false;
}
}
public String globalInfo() {
return "A tree is produced by assigning each individual the closest individual with better fitness. Connections with a distance above a certain threshold are cut. After that, each interconnected subtree forms a cluster.";
}
public String metricTipText() {
return "The metric to use during clustering.";
}
public InterfaceDistanceMetric getMetric() {
return metric;
}
public void setMetric(InterfaceDistanceMetric metric) {
this.metric = metric;
}
public String distThresholdTipText() {
return "In the non-adaptive case the absolute threshold below which clusters are connected.";
}
public double getDistThreshold() {
return absoluteDistThreshold;
}
public void setDistThreshold(double distThreshold) {
this.absoluteDistThreshold = distThreshold;
}
public String minimumGroupSizeTipText() {
return "Minimum group size that makes an own cluster.";
}
public int getMinimumGroupSize() {
return minimumGroupSize;
}
public void setMinimumGroupSize(int minimumGroupSize) {
this.minimumGroupSize = minimumGroupSize;
}
public String comparatorTipText() {
return "Define the comparator by which the population is sorted before clustering.";
}
public AbstractEAIndividualComparator getComparator() {
return comparator;
}
// public void setComparator(AbstractEAIndividualComparator comparator) {
// this.comparator = comparator;
// }
public String adaptiveThresholdTipText() {
return "Activate adaptive threshold which is calculated from mean distance in the population and a constant factor.";
}
public boolean isAdaptiveThreshold() {
return thresholdMultipleOfMeanDist;
}
public void setAdaptiveThreshold(boolean thresholdMultipleOfMeanDist) {
this.thresholdMultipleOfMeanDist = thresholdMultipleOfMeanDist;
GenericObjectEditor.setHideProperty(this.getClass(), "meanDistFactor", !thresholdMultipleOfMeanDist);
GenericObjectEditor.setHideProperty(this.getClass(), "distThreshold", thresholdMultipleOfMeanDist);
}
public String meanDistFactorTipText() {
return "Factor producing the distance threshold from population mean distance.";
}
public double getMeanDistFactor() {
return meanDistFactor;
}
public void setMeanDistFactor(double meanDistFactor) {
this.meanDistFactor = meanDistFactor;
}
public String testConvergingSpeciesOnBestOnlyTipText() {
return "Only the best individuals may be compared when testing whether to merge two species.";
}
public boolean isTestConvergingSpeciesOnBestOnly() {
return testConvergingSpeciesOnBestOnly;
}
public void SetTestConvergingSpeciesOnBestOnly(
boolean testConvergingSpeciesOnBestOnly) {
this.testConvergingSpeciesOnBestOnly = testConvergingSpeciesOnBestOnly;
}
}

39
src/eva2/server/go/operators/cluster/ClusteringXMeans.java
View File

@@ -1,5 +1,7 @@
package eva2.server.go.operators.cluster;
import java.util.Arrays;
import eva2.gui.Chart2DDPointIconCircle;
import eva2.gui.Chart2DDPointIconText;
import eva2.gui.GraphPointSet;
@@ -53,7 +55,7 @@ public class ClusteringXMeans implements InterfaceClustering, java.io.Serializab
* @param pop The population of individuals that is to be clustered.
* @return Population[]
*/
public Population[] cluster(Population pop) {
public Population[] cluster(Population pop, Population referencePop) {
ClusteringKMeans kmeans = new ClusteringKMeans();
Population[][] tmpResults = new Population[this.m_MaxK][];
double[][][] tmpC = new double[this.m_MaxK][][];
@@ -68,7 +70,7 @@ public class ClusteringXMeans implements InterfaceClustering, java.io.Serializab
for (int i = 1; i < this.m_MaxK; i++) {
kmeans.setUseSearchSpace(this.m_UseSearchSpace);
kmeans.setK(i+1);
tmpResults[i] = kmeans.cluster(pop);
tmpResults[i] = kmeans.cluster(pop, (Population)null);
tmpC[i] = kmeans.getC();
}
@@ -221,21 +223,28 @@ public class ClusteringXMeans implements InterfaceClustering, java.io.Serializab
* @param species2 The second species.
* @return True if species converge, else False.
*/
public boolean mergingSpecies(Population species1, Population species2) {
public boolean mergingSpecies(Population species1, Population species2, Population referencePop) {
// @todo i could use the BIC metric from X-means to calculate this
return false;
}
/** This method decides if a unclustered individual belongs to an already established species.
* @param indy A unclustered individual.
* @param species A species.
* @return True or False.
*/
public boolean belongsToSpecies(AbstractEAIndividual indy, Population species) {
// @todo perhaps the same as in convergingSpecies
return false;
}
// /** This method decides if a unclustered individual belongs to an already established species.
// * @param indy A unclustered individual.
// * @param species A species.
// * @return True or False.
// */
// public boolean belongsToSpecies(AbstractEAIndividual indy, Population species, Population pop) {
// // @todo perhaps the same as in convergingSpecies
// return false;
// }
public int[] associateLoners(Population loners, Population[] species, Population referencePop) {
int[] res=new int[loners.size()];
System.err.println("Warning, associateLoners not implemented for " + this.getClass());
Arrays.fill(res, -1);
return res;
}
/** This method allows you to recieve the c centroids
* @return The centroids
*/
@@ -322,7 +331,7 @@ public class ClusteringXMeans implements InterfaceClustering, java.io.Serializab
} else {
f1.initPopulation(pop);
}
ckm.cluster(pop);
ckm.cluster(pop, (Population)null);
}
@@ -370,4 +379,8 @@ public class ClusteringXMeans implements InterfaceClustering, java.io.Serializab
public String useSearchSpaceTipText() {
return "Toggel between search/objective space distance.";
}
public String initClustering(Population pop) {
return null;
}
}

35
src/eva2/server/go/operators/cluster/InterfaceClustering.java
View File

@@ -33,19 +33,46 @@ public interface InterfaceClustering {
* @param pop The population of individuals that is to be clustered.
* @return Population[]
*/
public Population[] cluster(Population pop);
public Population[] cluster(Population pop, Population referenceSet);
/** This method allows you to decide if two species are to be merged regarding this clustering algorithm.
* @param species1 The first species.
* @param species2 The second species.
* @return True if species converge, else False.
*/
public boolean mergingSpecies(Population species1, Population species2);
public boolean mergingSpecies(Population species1, Population species2, Population referenceSet);
/** This method decides if an unclustered individual belongs to an already established species.
/**
* Do some pre-calculations on a population for clustering. If additional population data
* is set, return the associated key, otherwise null.
*
* @param pop
*/
public String initClustering(Population pop);
/**
* This method decides if an unclustered individual belongs to an already established species.
* For some clustering methods this can only be decided in reference to the complete population.
*
* @param indy A unclustered individual.
* @param species A species.
* @param pop The complete population as a reference.
* @return True or False.
*/
public boolean belongsToSpecies(AbstractEAIndividual indy, Population species);
//Removed since for some clustering methods its not feasible to associate loners sequentially. Instead, a whole set of
// lone individuals can now be associated to a given set of clusters
//public boolean belongsToSpecies(AbstractEAIndividual indy, Population species);
/**
* Try to associate a set of loners with a given set of species. Return a list
* of indices assigning loner i with species j for all loners. If no species can
* be associated, -1 is returned as individual entry.
* Note that the last cluster threshold is used which may have depended on the last
* generation.
*
* @param loners
* @param species
* @return associative list matching loners to species.
*/
public int[] associateLoners(Population loners, Population[] species, Population referenceSet);
}

8
src/eva2/server/go/operators/distancemetric/IndividualDataMetric.java
View File

@@ -5,6 +5,7 @@ import java.io.Serializable;
import eva2.server.go.individuals.AbstractEAIndividual;
import eva2.server.go.individuals.InterfaceDataTypeDouble;
import eva2.server.go.strategies.ParticleSwarmOptimization;
import eva2.tools.EVAERROR;
/**
* Define a metric on data stored within individuals, such as the personal best position
@@ -45,7 +46,12 @@ public class IndividualDataMetric implements InterfaceDistanceMetric, Serializab
double[][] range2 = ((InterfaceDataTypeDouble)indy2).getDoubleRange();
return EuclideanMetric.normedEuclideanDistance((double[])data1, range1, (double[])data2, range2);
} else return EuclideanMetric.euclideanDistance((double[])data1, (double[])data2);
} else throw new RuntimeException("Error, invalid key data, double array required by " + this.getClass().getName());
} else {
EVAERROR.errorMsgOnce("Error, invalid key data, double array required by " + this.getClass().getName());
EVAERROR.errorMsgOnce("Using PhenotypeMetric as Backup...");
return (new PhenotypeMetric().distance(indy1, indy2));
// throw new RuntimeException("Invalid data key, double array required by " + this.getClass().getName());
}
}
}

4
src/eva2/server/go/operators/fitnessmodifier/FitnessAdaptiveClustering.java
View File

@@ -42,9 +42,9 @@ public class FitnessAdaptiveClustering implements java.io.Serializable, Interfac
// also note that if all individual achieve equal fitness the sum will be zero
result[i] = data[i][x] -min + 0.1;
}
this.m_ClusteringAlgorithm.initClustering(population);
// Now search for clusters
Population[] ClusterResult = this.m_ClusteringAlgorithm.cluster(population);
Population[] ClusterResult = this.m_ClusteringAlgorithm.cluster(population, population);
Population cluster;
for (int i = 1; i < ClusterResult.length; i++) {
cluster = ClusterResult[i];

2
src/eva2/server/go/operators/migration/MOClusteringSeparation.java
View File

@@ -136,7 +136,7 @@ public class MOClusteringSeparation implements InterfaceMigration, java.io.Seria
// first set the K to the K-Means
this.m_KMeans.setK(islands.length);
this.m_KMeans.cluster(toCluster);
this.m_KMeans.cluster(toCluster, (Population)null);
double[][] c = this.m_KMeans.getC();
newIPOP = this.m_KMeans.cluster(collector, c);

2
src/eva2/server/go/operators/migration/MOXMeansSeparation.java
View File

@@ -131,7 +131,7 @@ public class MOXMeansSeparation implements InterfaceMigration, java.io.Serializa
// first set the K to the K-Means
this.m_XMeans.setMaxK(islands.length);
this.m_XMeans.cluster(toCluster);
this.m_XMeans.cluster(toCluster, (Population)null);
double[][] c = this.m_XMeans.getC();
//@todo Hier muss ich mal denk machen und weniger click...
newIPOP = this.m_XMeans.cluster(collector, c);

7
src/eva2/server/go/operators/postprocess/PostProcess.java
View File

@@ -173,7 +173,8 @@ public class PostProcess {
//cluster the undifferentiated population
Population result = new Population(10);
result.setSameParams(pop);
Population[] clusters = clustering.cluster(pop);
clustering.initClustering(pop);
Population[] clusters = clustering.cluster(pop, null);
if (TRACE) {
System.out.println("found " + clusters.length + " clusters!");
int sum=0;
@@ -423,7 +424,7 @@ public class PostProcess {
ppRunnable = new OptimizerRunnable(OptimizerFactory.makeParams(nms, pop, problem, 0, term), true);
// as nms creates a new population and has already evaluated them, send a signal to stats
ppRunnable.getStats().createNextGenerationPerformed(nms.getPopulation(), null);
ppRunnable.getStats().createNextGenerationPerformed(nms.getPopulation(), nms, null);
runPP();
@@ -464,7 +465,7 @@ public class PostProcess {
pop.SetFunctionCalls(baseEvals);
ppRunnable = new OptimizerRunnable(cmaParams, true);
ppRunnable.getStats().createNextGenerationPerformed(cmaParams.getOptimizer().getPopulation(), null);
ppRunnable.getStats().createNextGenerationPerformed(cmaParams.getOptimizer().getPopulation(), cmaParams.getOptimizer(), null);
runPP();
pop.clear();

235
src/eva2/server/go/populations/Population.java
View File

@@ -69,6 +69,7 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
// a sorted queue (for efficiency)
transient private ArrayList<AbstractEAIndividual> sortedArr = null;
private int lastFitCrit = -1;
private transient static boolean TRACE = false;
// remember when the last evaluation was performed
// private Pair<Integer,Integer> evaluationTimeHashes = null;
@@ -76,6 +77,7 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
// private int evaluationTimeModCount = -1;
public Population() {
if (TRACE) System.err.println("TRACING POP");
}
/**
@@ -86,10 +88,17 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
*/
public Population(int initialCapacity) {
super(initialCapacity);
if (TRACE) System.err.println("TRACING POP");
setTargetSize(initialCapacity);
}
/**
* Clones parameters, individuals, history and archive.
*
* @param population
*/
public Population(Population population) {
if (TRACE) System.err.println("TRACING POP");
setSameParams(population);
for (int i = 0; i < population.size(); i++) {
if (population.get(i) != null)
@@ -97,10 +106,49 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
}
copyHistAndArchive(population);
}
public void copyHistAndArchive(Population population) {
public static Population makePopFromList(List<AbstractEAIndividual> indies) {
Population pop = new Population(indies.size());
if (TRACE) System.err.println("TRACING POP");
pop.setTargetSize(indies.size());
for (AbstractEAIndividual indy : indies) {
pop.add(indy);
}
return pop;
}
/**
* Create a new population from a solution set by merging
* both current population and solutions from the set.
*
* @param allSolutions
*/
public Population(InterfaceSolutionSet allSolutions) {
this(allSolutions.getCurrentPopulation().size()+allSolutions.getSolutions().size());
if (TRACE) System.err.println("TRACING POP");
addPopulation(allSolutions.getCurrentPopulation());
HashMap<Long, Integer> checkCols = new HashMap<Long, Integer>(size());
for (int i=0; i<size(); i++) {
checkCols.put(getEAIndividual(i).getIndyID(), 1);
}
Population sols = allSolutions.getSolutions();
for (int i=0; i<sols.size();i++) {
// addPopulation(allSolutions.getSolutions());
if (!checkCols.containsKey(sols.getEAIndividual(i).getIndyID())) add(sols.getEAIndividual(i));
}
}
public void copyHistAndArchive(Population population) {
if (population.m_Archive != null) this.m_Archive = (Population)population.m_Archive.clone();
if (population.m_History != null) this.m_History = (ArrayList)population.m_History.clone();
if (population.m_History != null) this.m_History = (ArrayList<AbstractEAIndividual>)population.m_History.clone();
if (population.additionalPopData!=null) {
this.additionalPopData = (HashMap<String, Object>)additionalPopData.clone();
if (population.additionalPopData.size()>0) {
for (String key : population.additionalPopData.keySet()) {
additionalPopData.put(key, population.additionalPopData.get(key));
}
}
}
}
/**
@@ -173,7 +221,7 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
* have been inited by a problem
*/
public void init() {
this.m_History = new ArrayList();
this.m_History = new ArrayList<AbstractEAIndividual>();
this.m_Generation = 0;
this.m_FunctionCalls = 0;
// evaluationTimeHashes = null;
@@ -302,8 +350,8 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
*/
protected void firePropertyChangedEvent(String name) {
if (listeners != null) {
for (Iterator iterator = listeners.iterator(); iterator.hasNext();) {
InterfacePopulationChangedEventListener listener = (InterfacePopulationChangedEventListener) iterator.next();
for (Iterator<InterfacePopulationChangedEventListener> iterator = listeners.iterator(); iterator.hasNext();) {
InterfacePopulationChangedEventListener listener = iterator.next();
if (listener!=null) listener.registerPopulationStateChanged(this, name);
}
}
@@ -390,17 +438,86 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
* Note: After this operation the target population size may be exceeded.
* @param pop The population that is to be added.
*/
public void addPopulation(Population pop) {
if (pop == null) return;
for (int i = 0; i < pop.size(); i++) {
AbstractEAIndividual indy = (AbstractEAIndividual)pop.get(i);
if (indy != null) {
this.add(indy);
public Population addPopulation(Population pop) {
if (pop != null) {
for (int i = 0; i < pop.size(); i++) {
AbstractEAIndividual indy = (AbstractEAIndividual)pop.get(i);
if (indy != null) {
this.add(indy);
}
}
}
return this;
}
/**
* Fill the population up to the given size with random elements
* from the given population.
*
* @param upTo target size of the population
* @param fromPop The population that is to be added.
*
*/
public boolean fillWithRandom(int upTo, Population fromPop) {
if (upTo <= this.size()) return true;
else if (fromPop==null || (fromPop.size()<1)) return false;
else {
int[] perm = RNG.randomPerm(fromPop.size());
int i=0;
while ((i<perm.length) && (this.size()<upTo)) { // until instance is filled or no more indys can be selected
AbstractEAIndividual indy = (AbstractEAIndividual)fromPop.get(perm[i]);
// System.out.println("checking " + indy.getStringRepresentation());
if ((indy != null) && (!containsByPosition(indy))) {
this.add(indy);
} else {
// System.out.println(indy.getStringRepresentation() + " was contained!");
}
i++;
}
if (size()==upTo) return true;
else return false;
}
}
/**
* Find a list of pairs of indices within the population at which individuals with equal
* positions can be found.
* @return
*/
public List<Pair<Integer,Integer>> findSamePositions() {
ArrayList<Pair<Integer,Integer>> dupes = new ArrayList<Pair<Integer,Integer>>();
for (int i=0; i<size()-1; i++) {
int nextIndex = indexByPosition(i+1, getEAIndividual(i));
if (nextIndex >= 0) dupes.add(new Pair<Integer,Integer>(i, nextIndex));
}
return dupes;
}
/**
* Return true if an individual with equal position is contained within the population.
*
* @param indy
* @return
*/
public boolean containsByPosition(AbstractEAIndividual indy) {
return indexByPosition(0,indy)>=0;
}
/**
* Return the index of the first individual which has an equal position or -1.
*
* @param startIndex the first index to start the search
* @param indy
* @return the index of the first individual which has an equal position or -1
*/
public int indexByPosition(int startIndex, AbstractEAIndividual indy) {
for (int i=startIndex; i<size(); i++) {
if (Arrays.equals(AbstractEAIndividual.getDoublePositionShallow(indy), AbstractEAIndividual.getDoublePositionShallow(getEAIndividual(i)))) return i;
}
return -1;
}
/**
* Resets the fitnes to the maximum possible value for the given individual.
*
* @param indy an individual whose fitness will be reset
@@ -799,6 +916,7 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
* @return
*/
public ArrayList<AbstractEAIndividual> getSorted(AbstractEAIndividualComparator comp) {
if (super.size()==0) return new ArrayList<AbstractEAIndividual>();
PriorityQueue<AbstractEAIndividual> sQueue = new PriorityQueue<AbstractEAIndividual>(super.size(), comp);
for (int i = 0; i < super.size(); i++) {
AbstractEAIndividual indy = getEAIndividual(i);
@@ -1132,7 +1250,8 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
}
/**
* ArrayList does not increase the modCount in set. Why???
* ArrayList does not increase the modCount in set. Maybe because it is not
* seen as "structural change"?
*/
public Object set(int index, Object element) {
Object prev = super.set(index, element);
@@ -1250,10 +1369,32 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
* @return the average, minimal and maximal mean distance of individuals in an array of three
*/
public double[] getPopulationMeasures(InterfaceDistanceMetric metric) {
double d;
// Integer lastMeasuresModCount = (Integer)getData(lastPopMeasuresFlagKey);
// if (lastMeasuresModCount!=null && (lastMeasuresModCount==modCount)) {
// double[] measures = (double[])getData(lastPopMeasuresDatKey);
// if (TRACE ) {
// double[] verify = calcPopulationMeasures(metric);
// if (Mathematics.dist(measures, verify, 2)>1e-10) {
// System.out.println(getStringRepresentation());
// System.err.println("Warning, invalid measures!!!");
//
// }
// }
// return measures;
// }
double[] res = calcPopulationMeasures(metric);
// putData(lastPopMeasuresFlagKey, new Integer(modCount));
// putData(lastPopMeasuresDatKey, res);
// System.out.println(getStringRepresentation());
// System.out.println("0-1-dist: " + BeanInspector.toString(metric.distance((AbstractEAIndividual)this.get(0), (AbstractEAIndividual)this.get(1))));
return res;
}
private double[] calcPopulationMeasures(InterfaceDistanceMetric metric) {
double d;
double[] res = new double[3];
double distSum = 0.;
double meanDist = 0.;
double maxDist = Double.MIN_VALUE;
double minDist = Double.MAX_VALUE;
@@ -1262,21 +1403,20 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
if (metric == null) d = EuclideanMetric.euclideanDistance(AbstractEAIndividual.getDoublePositionShallow(getEAIndividual(i)),
AbstractEAIndividual.getDoublePositionShallow(getEAIndividual(j)));
else d = metric.distance((AbstractEAIndividual)this.get(i), (AbstractEAIndividual)this.get(j));
distSum += d;
meanDist += d;
if (d < minDist) minDist = d;
if (d > maxDist) maxDist = d;
}
}
res[1] = minDist;
res[2] = maxDist;
if (this.size() > 1) res[0] = distSum / (this.size() * (this.size()-1) / 2);
if (this.size() > 1) res[0] = meanDist / (this.size() * (this.size()-1) / 2);
else { // only one indy?
res[1]=0;
res[2]=0;
}
// System.out.println("0-1-dist: " + BeanInspector.toString(metric.distance((AbstractEAIndividual)this.get(0), (AbstractEAIndividual)this.get(1))));
return res;
}
return res;
}
/**
* Returns the average, minimal and maximal individual fitness and std dev. for the population in the given criterion.
@@ -1342,7 +1482,30 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
}
return new Pair<Integer,Double>(sel,dist);
}
/**
* Search for the closest (farthest) individual to the given individual.
* Return a Pair of the individuals index and distance.
* If the population is empty, a Pair of (-1,-1) is returned.
*
* @param pos
* @param pop
* @param closestOrFarthest if true, the closest individual is retrieved, otherwise the farthest
* @return
*/
public static Pair<Integer,Double> getClosestFarthestIndy(AbstractEAIndividual refIndy, Population pop, InterfaceDistanceMetric metric, boolean closestOrFarthest) {
double dist = -1.;
int sel=-1;
for (int i = 0; i < pop.size(); i++) {
double curDist = metric.distance(refIndy, pop.getEAIndividual(i));
if ((dist<0) || (!closestOrFarthest && (dist < curDist))
|| (closestOrFarthest && (dist > curDist))) {
dist = curDist;
sel = i;
}
}
return new Pair<Integer,Double>(sel,dist);
}
/**
* Calculate the average position of the population.
*
@@ -1574,6 +1737,36 @@ public class Population extends ArrayList implements PopulationInterface, Clonea
// if (evaluationTimeHashes == null) return false;
// else return ((hashes.head().equals(evaluationTimeHashes.head())) && (hashes.tail().equals(evaluationTimeHashes.tail())) && (evaluationTimeModCount == modCount));
// }
/**
* Add the population data of a given population to this instance.
* Note that collisions are not checked!
*/
public void addDataFromPopulation(Population pop) {
if (pop.additionalPopData!=null) {
Set<String> keys = pop.additionalPopData.keySet();
for (String key : keys) {
Object earlierDat = this.getData(key);
if (earlierDat!=null && !(earlierDat.equals(pop.getData(key)))) System.err.println("Warning: Population already contained data keyed by " + key + ", overwriting data " + earlierDat + " with " + pop.getData(key));
this.putData(key, pop.getData(key));
}
}
}
/**
* Returns true if the given individual is member of this population,
* where the comparison is done by individual ID.
*
* @param indy
* @return
*/
public boolean isMemberByID(AbstractEAIndividual indy) {
for (int i=0; i<size(); i++) {
if (getEAIndividual(i).getIndyID()==indy.getIndyID()) {
return true;
}
}
return false;
}
/**
* Return true if the targeted size of the population has been reached.

1002
src/eva2/server/go/strategies/ClusterBasedNichingEA.java
View File

File diff suppressed because it is too large Load Diff

4
src/eva2/server/go/strategies/DifferentialEvolution.java
View File

@@ -250,8 +250,8 @@ public class DifferentialEvolution implements InterfaceOptimizer, java.io.Serial
indy = (AbstractEAIndividual)(pop.getEAIndividual(parentIndex)).getClone();
esIndy = (InterfaceDataTypeDouble)indy;
} catch (java.lang.ClassCastException e) {
EVAERROR.errorMsgOnce("Differential Evolution currently requires InterfaceESIndividual as basic data type!");
return (AbstractEAIndividual)((AbstractEAIndividual)pop.get(RNG.randomInt(0, pop.size()-1))).getClone();
throw new RuntimeException("Differential Evolution currently requires InterfaceESIndividual as basic data type!");
// return (AbstractEAIndividual)((AbstractEAIndividual)pop.get(RNG.randomInt(0, pop.size()-1))).getClone();
}
double[] nX, vX, oX;
oX = esIndy.getDoubleData();

7
src/eva2/server/go/strategies/EvolutionStrategies.java
View File

@@ -239,7 +239,7 @@ public class EvolutionStrategies implements InterfaceOptimizer, java.io.Serializ
*/
protected Population getReplacePop(Population nextGeneration) {
if (forceOrigPopSize && (origPopSize > 0) && (origPopSize < nextGeneration.size())) {
// this is especially for CBN:
// this is especially for CBN: earlier selection to immediately reduce the size of mu+lambda to lambda
this.m_EnvironmentSelection.prepareSelection(nextGeneration);
Population tmpPop = (Population)nextGeneration.clone();
nextGeneration.clear();
@@ -368,8 +368,9 @@ public class EvolutionStrategies implements InterfaceOptimizer, java.io.Serializ
System.err.println("Invalid mu/lambda ratio! Setting mu=lambda="+m_Mu);
this.m_Lambda = this.m_Mu;
}
if (this.m_UsePlusStrategy) this.m_Population.setTargetSize(this.m_Mu + this.m_Lambda);
else this.m_Population.setTargetSize(this.m_Lambda);
// if (this.m_UsePlusStrategy) this.m_Population.setTargetSize(this.m_Mu + this.m_Lambda);
// else this.m_Population.setTargetSize(this.m_Lambda);
this.m_Population.setTargetSize(this.m_Lambda);
origPopSize=m_Population.getTargetSize();
}

32
src/eva2/server/go/strategies/InterfaceSpeciesAware.java Normal file
View File

@@ -0,0 +1,32 @@
package eva2.server.go.strategies;
import eva2.server.go.populations.Population;
/**
* An interface for optimizers which are to be notified in case of species
* based optimization; namely merging and split events.
*
* @author mkron
*
*/
public interface InterfaceSpeciesAware {
// these can be used to tag a population as explorer or local search population.
public final static String populationTagKey="specAwarePopulationTag";
public final static Integer explorerPopTag=23;
public final static Integer localPopTag=42;
/**
* Two species have been merged to the first one.
* @param p1
* @param p2
*/
public void mergeToFirstPopulationEvent(Population p1, Population p2);
/**
* Notify that a split has occurred separating p2 from p1.
*
* @param p1
* @param p2
*/
public void splitFromFirst(Population p1, Population p2);
}

4
src/eva2/server/go/strategies/ParticleSwarmOptimization.java
View File

@@ -1201,7 +1201,7 @@ public class ParticleSwarmOptimization implements InterfaceOptimizer, java.io.Se
}
public void optimize() {
// System.out.println(">>> " + m_Population.getStringRepresentation());
startOptimize();
// Update the individuals
@@ -1216,7 +1216,7 @@ public class ParticleSwarmOptimization implements InterfaceOptimizer, java.io.Se
// log the best individual of the population
logBestIndividual();
// System.out.println(">>> " + m_Population.getBestEAIndividual().getStringRepresentation());
// System.out.println("<<< " + m_Population.getStringRepresentation());
// if (doLocalSearch && (m_Population.getGeneration()%localSearchGens==0)) {
//// System.out.println("Local search at gen "+m_Population.getGeneration());

1
src/eva2/server/modules/Processor.java
View File

@@ -343,6 +343,7 @@ public class Processor extends Thread implements InterfaceProcessor, InterfacePo
if (this.goParams.getOptimizer() instanceof InterfaceAdditionalPopulationInformer) informerList.add(this.goParams.getOptimizer());
m_Statistics.createNextGenerationPerformed(
(PopulationInterface)this.goParams.getOptimizer().getPopulation(),
this.goParams.getOptimizer(),
informerList);
if (m_ListenerModule != null) {
m_ListenerModule.updateProgress(

18
src/eva2/server/stat/AbstractStatistics.java
View File

@@ -14,8 +14,10 @@ import eva2.server.go.IndividualInterface;
import eva2.server.go.PopulationInterface;
import eva2.server.go.individuals.AbstractEAIndividual;
import eva2.server.go.operators.distancemetric.InterfaceDistanceMetric;
import eva2.server.go.populations.InterfaceSolutionSet;
import eva2.server.go.populations.Population;
import eva2.server.go.problems.InterfaceAdditionalPopulationInformer;
import eva2.server.go.strategies.InterfaceOptimizer;
import eva2.tools.Mathematics;
import eva2.tools.Pair;
@@ -77,7 +79,7 @@ public abstract class AbstractStatistics implements InterfaceTextListener, Inter
private ArrayList<InterfaceTextListener> textListeners;
private List<InterfaceAdditionalPopulationInformer> lastInformerList = null;
private PopulationInterface lastPop = null;
private PopulationInterface lastSols = null;
public AbstractStatistics() {
firstPlot = true;
@@ -173,7 +175,7 @@ public abstract class AbstractStatistics implements InterfaceTextListener, Inter
currentBestFeasibleFit=null;
bestOfRunFeasibleIndy = null;
lastInformerList = null;
lastPop = null;
lastSols = null;
runIterCnt = 0;
if (printRunIntroVerbosity()) printToTextListener("\n****** Multirun "+runNumber);
if (params != null) {
@@ -578,8 +580,7 @@ public abstract class AbstractStatistics implements InterfaceTextListener, Inter
* @param pop
*/
private void updateLastAdditionalInfo() {
// TODO Auto-generated method stub
Double[] lastVals = appendAdditionalInfo(lastInformerList, lastPop, null);
Double[] lastVals = appendAdditionalInfo(lastInformerList, lastSols, null);
lastAdditionalInfoSums = updateAdditionalInfo(lastAdditionalInfoSums, lastVals);
}
@@ -602,9 +603,8 @@ public abstract class AbstractStatistics implements InterfaceTextListener, Inter
*
*/
public synchronized void createNextGenerationPerformed(PopulationInterface
pop, List<InterfaceAdditionalPopulationInformer> informerList) {
pop, InterfaceOptimizer opt, List<InterfaceAdditionalPopulationInformer> informerList) {
lastInformerList = informerList;
lastPop = pop;
if (firstPlot) {
initPlots(m_StatsParams.getPlotDescriptions());
// if (doTextOutput()) printToTextListener(getOutputHeader(informer, pop)+'\n');
@@ -676,7 +676,7 @@ public abstract class AbstractStatistics implements InterfaceTextListener, Inter
meanCollection.add(means);
} else {
if (meanCollection.size()<=runIterCnt) {// bad case!
// may happen for dynamic pop-sizes, e.g. in Tribe, when runs do not necessarily send the
// may happen for dynamic pop-sizes, e.g. in Tribes, when runs do not necessarily send the
// "generation performed" event the same number of times.
// thus: dont do an update for events that are "too late"
means = null;
@@ -686,8 +686,10 @@ public abstract class AbstractStatistics implements InterfaceTextListener, Inter
}
// meanCollection.set(pop.getGenerations()-1, means);
lastSols = (opt!=null) ? new Population(opt.getAllSolutions()) : pop;
if (doTextOutput()) {
Pair<String,Double[]> addInfo = getOutputLine(informerList, pop);
Pair<String,Double[]> addInfo = getOutputLine(informerList, lastSols);
if (printLineByVerbosity(runIterCnt)) printToTextListener(addInfo.head()+'\n');
// updateAdditionalInfo(addInfo.tail());
if (addInfo.tail()!=null) {

3
src/eva2/server/stat/InterfaceStatistics.java
View File

@@ -17,6 +17,7 @@ import java.util.List;
import eva2.server.go.IndividualInterface;
import eva2.server.go.PopulationInterface;
import eva2.server.go.problems.InterfaceAdditionalPopulationInformer;
import eva2.server.go.strategies.InterfaceOptimizer;
/*==========================================================================*
* INTERFACE DECLARATION
*==========================================================================*/
@@ -35,7 +36,7 @@ public interface InterfaceStatistics {
public void addTextListener(InterfaceTextListener listener);
public boolean removeTextListener(InterfaceTextListener listener);
public void printToTextListener(String s);
public void createNextGenerationPerformed(PopulationInterface Pop, List<InterfaceAdditionalPopulationInformer> informerList);
public void createNextGenerationPerformed(PopulationInterface Pop, InterfaceOptimizer opt, List<InterfaceAdditionalPopulationInformer> informerList);
public void createNextGenerationPerformed(double[] bestfit,double[] worstfit,int calls);
public InterfaceStatisticsParameter getStatisticsParameter(); // called from moduleadapter
public IndividualInterface getRunBestSolution(); // return the best fitness of the last run (may not be equal to the last population)

3
src/eva2/server/stat/StatisticsDummy.java
View File

@@ -6,6 +6,7 @@ import eva2.server.go.IndividualInterface;
import eva2.server.go.PopulationInterface;
import eva2.server.go.individuals.AbstractEAIndividual;
import eva2.server.go.problems.InterfaceAdditionalPopulationInformer;
import eva2.server.go.strategies.InterfaceOptimizer;
/**
* This may be given to a Processor if no further stats are required. It speeds up
@@ -36,7 +37,7 @@ public class StatisticsDummy implements InterfaceStatistics, InterfaceTextListen
System.err.println("addTextListener not provided!");
}
public void createNextGenerationPerformed(PopulationInterface pop,
public void createNextGenerationPerformed(PopulationInterface pop, InterfaceOptimizer opt,
List<InterfaceAdditionalPopulationInformer> informerList) {
bestCurrentIndividual = (AbstractEAIndividual)pop.getBestIndividual();
if ((bestIndividualAllover == null) || (AbstractStatistics.secondIsBetter(bestIndividualAllover, bestCurrentIndividual))) {

19
src/eva2/server/stat/StatisticsWithGUI.java
View File

@@ -247,23 +247,12 @@ public class StatisticsWithGUI extends AbstractStatistics implements Serializabl
int subGraph=0;
if (doPlotCurrentBest) plotFitnessPoint(0, subGraph++, functionCalls, currentBestFit[0]);
if (doPlotRunBest) plotFitnessPoint(0, subGraph++, functionCalls, bestOfRunIndy.getFitness()[0]);
if (doPlotWorst) {// schlechteste Fitness plotten
if (currentWorstFit == null) {
System.err.println("m_WorstFitness==null in plotStatisticsPerformed");
return;
}
plotFitnessPoint(0, subGraph++ , functionCalls, currentWorstFit[0]);
}
if (doPlotWorst) plotFitnessPoint(0, subGraph++ , functionCalls, currentWorstFit[0]);
if (doPlotAvgDist) plotFitnessPoint(0, subGraph++, functionCalls, avgPopDist);
if (doPlotMaxPopDist) plotFitnessPoint(0, subGraph++, functionCalls, maxPopDist);
if (doPlotCurBestFeasible && currentBestFeasibleFit!=null) {
plotFitnessPoint(0, subGraph++, functionCalls, currentBestFeasibleFit[0]);
}
if (doPlotRunBestFeasible && bestOfRunFeasibleIndy!=null) {
plotFitnessPoint(0, subGraph++, functionCalls, bestOfRunFeasibleIndy.getFitness()[0]);
}
if (doPlotCurBestFeasible && currentBestFeasibleFit!=null) plotFitnessPoint(0, subGraph++, functionCalls, currentBestFeasibleFit[0]);
if (doPlotRunBestFeasible && bestOfRunFeasibleIndy!=null) plotFitnessPoint(0, subGraph++, functionCalls, bestOfRunFeasibleIndy.getFitness()[0]);
}
}