package jsat.classifiers;
import java.util.PriorityQueue;
/**
* Decision Directed Acyclic Graph (DDAG) classifier. DDAG extends
* binary decision classifiers into multi-class decision classifiers. Unlike
* {@link OneVSOne}, DDAG results are hard classification decisions, and will
* not give probabilistic estimates. Accuracy is often very similar, but DDAG
* classification speed can be significantly faster, as it does not evaluate all
* possible combinations of classifiers.
*
* @author Edward Raff
*/
public class DDAG extends OneVSOne
{
private static final long serialVersionUID = -9109002614319657144L;
/**
* Creates a new DDAG classifier to extend a binary classifier to handle multi-class problems.
* @param baseClassifier the binary classifier to extend
* @param concurrentTrain true to have training of individual
* classifiers occur in parallel, false to have them use their
* native parallel training method.
*/
public DDAG(Classifier baseClassifier, boolean concurrentTrain)
{
super(baseClassifier, concurrentTrain);
}
/**
* Creates a new DDAG classifier to extend a binary classifier to handle multi-class problems.
* @param baseClassifier the binary classifier to extend
*/
public DDAG(Classifier baseClassifier)
{
super(baseClassifier);
}
@Override
public CategoricalResults classify(DataPoint data)
{
CategoricalResults cr = new CategoricalResults(predicting.getNumOfCategories());
//Use a priority que so that we always pick the two lowest value class labels, makes indexing into the oneVsOne array simple
PriorityQueue options = new PriorityQueue(predicting.getNumOfCategories());
for(int i = 0; i < cr.size(); i++)
options.add(i);
CategoricalResults subRes;
int c1, c2;
//We will now loop through and repeatedly pick two combinations, and eliminate the loser, until there is one winer
while(options.size() > 1)
{
c1 = options.poll();
c2 = options.poll();
subRes = oneVone[c1][c2-c1-1].classify(data);
if(subRes.mostLikely() == 0)//c1 wins, c2 no longer a candidate
options.add(c1);
else//c2 wins, c1 no onger a candidate
options.add(c2);
}
cr.setProb(options.peek(), 1.0);
return cr;
}
@Override
public DDAG clone()
{
DDAG clone = new DDAG(baseClassifier.clone(), isConcurrentTraining());
if (oneVone != null)
{
clone.oneVone = new Classifier[oneVone.length][];
for (int i = 0; i < oneVone.length; i++)
{
clone.oneVone[i] = new Classifier[oneVone[i].length];
for (int j = 0; j < oneVone[i].length; j++)
clone.oneVone[i][j] = oneVone[i][j].clone();
}
}
if(predicting != null)
clone.predicting = predicting.clone();
return clone;
}
}
Ads help maintain this website.