relational.classifier
Class Iterative

java.lang.Object
  extended byrelational.classifier.Iterative

public class Iterative
extends java.lang.Object

Class contains Collective Inference methods


Field Summary
 java.util.Hashtable classified
           
 
Constructor Summary
Iterative()
           
 
Method Summary
 void relaxationLabeling(weka.classifiers.Classifier prncl, weka.core.Instances test, java.util.HashMap id2ClassProb, edu.uci.ics.jung.graph.impl.SparseGraph graph, int maxiterations, java.util.HashMap weightsMap, java.util.HashMap probabilities, java.lang.String type)
          Iterative method for Relational Neighbor Classifier, it uses a heterogenous, directed, weighted, bipartite graph Only the first classe of an instance is considered
 void relaxationLabeling(relational.classifier.ProbRN2Hop prncl, weka.core.Instances test, java.util.HashMap id2ClassProb, edu.uci.ics.jung.graph.impl.SparseGraph graph, int maxiterations, int numNeighbors, java.util.HashMap weightsMap, java.util.HashMap probabilities, java.lang.String type)
          Iterative Method for PRN2Hop, uses a bipartite graph Only the first class of an instance is considered
 void relaxationLabeling(ProbRNMultiHop prncl, weka.core.Instances test, java.util.HashMap id2ClassProb, edu.uci.ics.jung.graph.impl.SparseGraph graph, int maxiterations, double minWeight, java.util.HashMap weightsMap, java.util.HashMap normWeights, java.util.HashMap probabilities, java.lang.String type, int numNeighbors, int hops)
          Iterative method for PRNMultiHop, uses a bipartite graph.
 void relaxationLabelingMultiLabel(ProbRNMultiHopMultiLabel prncl, weka.core.Instances test, java.util.HashMap id2ClassProb, edu.uci.ics.jung.graph.impl.SparseGraph graph, int label, java.util.HashMap expert, java.util.HashMap result, double th, double minWeight, int maxIterations, java.util.HashMap probabilities, int numNeighbors, java.util.HashMap weights, java.util.HashMap normWeights, java.lang.String type, int hops)
          Iterative method for PRNMultiHop All categories of an instance are considered (binary classification)
 void relaxationLabelingMultiLabel(relational.classifier.ProbRNMultiLabel2Hop prncl, weka.core.Instances test, java.util.HashMap id2ClassProb, edu.uci.ics.jung.graph.impl.SparseGraph graph, int label, java.util.HashMap expert, java.util.HashMap result, double th, int maxIterations, java.util.HashMap probabilities, int numNeighbors, java.util.HashMap weights, java.lang.String type)
          Iterative method for PRN2Hop.
 void relaxationLabelingMultiLabel(ProbRNMultiLabel prncl, weka.core.Instances test, java.util.HashMap id2ClassProb, edu.uci.ics.jung.graph.impl.SparseGraph graph, int label, java.util.HashMap expert, java.util.HashMap result, java.util.HashMap weights, double th, int maxIterations, java.util.HashMap probabilities, java.lang.String type)
          Iterative method for Probabilistic Relational Neighbor Classifier.
 void RNstar(RelationalNeighborClassifier cl, weka.core.Instances test, edu.uci.ics.jung.graph.impl.SparseGraph graph, java.util.HashMap id2Label, java.util.HashMap weights, java.util.HashMap probabilities, java.lang.String type)
          Iterative method for Relational Neighbor Classifier, it uses a homogenous, undirected, weighted graph Only the first class of an instance is considered
 void RNstarMultiLabel(RNCMultiLabel cl, weka.core.Instances test, java.util.HashMap id2Label, edu.uci.ics.jung.graph.impl.SparseGraph graph, int Class, java.util.HashMap expert, java.util.HashMap result, java.util.HashMap weights, double th, java.util.HashMap probabilities, java.lang.String type)
          Iterative method for Relational Neighbor Classifier, it uses a heterogenous, directed, weighted, bipartite graph All the classes of an instance are considered
 
Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Field Detail

classified

public java.util.Hashtable classified
Constructor Detail

Iterative

public Iterative()
Method Detail

RNstar

public void RNstar(RelationalNeighborClassifier cl,
                   weka.core.Instances test,
                   edu.uci.ics.jung.graph.impl.SparseGraph graph,
                   java.util.HashMap id2Label,
                   java.util.HashMap weights,
                   java.util.HashMap probabilities,
                   java.lang.String type)
            throws java.lang.Exception
Iterative method for Relational Neighbor Classifier, it uses a homogenous, undirected, weighted graph Only the first class of an instance is considered

Parameters:
cl - - classifier in inner loop
test - - test insatances
graph - - graph (jung)
id2Label - - initialization of test instances
weights - - the weights of the edges in the graph
probabilities - - resulting probabilities for each class and test instance
type - - indicates the type of problem (heterogenous or homogenous)
Throws:
java.lang.Exception

RNstarMultiLabel

public void RNstarMultiLabel(RNCMultiLabel cl,
                             weka.core.Instances test,
                             java.util.HashMap id2Label,
                             edu.uci.ics.jung.graph.impl.SparseGraph graph,
                             int Class,
                             java.util.HashMap expert,
                             java.util.HashMap result,
                             java.util.HashMap weights,
                             double th,
                             java.util.HashMap probabilities,
                             java.lang.String type)
                      throws java.lang.Exception
Iterative method for Relational Neighbor Classifier, it uses a heterogenous, directed, weighted, bipartite graph All the classes of an instance are considered

Parameters:
cl - - classifier in inner loop
test - - test insatances
graph - - graph (jung)
Class - - the category to be considered in binary classification
id2Label - - initialization of test instances
weights - - the weights of the edges in the graph
th - - threshold
probabilities - - resulting probabilities for each class and test instance
type - - indicates the type of problem (heterogenous or homogenous)
Throws:
java.lang.Exception

relaxationLabeling

public void relaxationLabeling(weka.classifiers.Classifier prncl,
                               weka.core.Instances test,
                               java.util.HashMap id2ClassProb,
                               edu.uci.ics.jung.graph.impl.SparseGraph graph,
                               int maxiterations,
                               java.util.HashMap weightsMap,
                               java.util.HashMap probabilities,
                               java.lang.String type)
                        throws java.lang.Exception
Iterative method for Relational Neighbor Classifier, it uses a heterogenous, directed, weighted, bipartite graph Only the first classe of an instance is considered

Parameters:
prncl - - classifier in inner loop
test - - test insatances
id2ClassProb - - initialization of test instances
graph - - graph (jung)
maxiterations - - the maximal number of iterations to be performed
weightsMap - - the weights of the edges in the graph
probabilities - - resulting probabilities for each class and test instance
type - - indicates the type of problem (heterogenous or homogenous)
Throws:
java.lang.Exception

relaxationLabeling

public void relaxationLabeling(relational.classifier.ProbRN2Hop prncl,
                               weka.core.Instances test,
                               java.util.HashMap id2ClassProb,
                               edu.uci.ics.jung.graph.impl.SparseGraph graph,
                               int maxiterations,
                               int numNeighbors,
                               java.util.HashMap weightsMap,
                               java.util.HashMap probabilities,
                               java.lang.String type)
                        throws java.lang.Exception
Iterative Method for PRN2Hop, uses a bipartite graph Only the first class of an instance is considered

Parameters:
prncl - - classifier in inner loop
test - - test insatances
id2ClassProb - - initialization of test instances
graph - - graph (jung)
maxiterations - - the maximal number of iterations to be performed
numNeighbors - - determines the number of neighbors for which additional indirect neighbors will be considered using PRN2Hop
weightsMap - - the weights of the edges in the graph
probabilities - - resulting probabilities for each class and test instance
type - - indicates the type of problem (heterogenous or homogenous)
Throws:
java.lang.Exception

relaxationLabeling

public void relaxationLabeling(ProbRNMultiHop prncl,
                               weka.core.Instances test,
                               java.util.HashMap id2ClassProb,
                               edu.uci.ics.jung.graph.impl.SparseGraph graph,
                               int maxiterations,
                               double minWeight,
                               java.util.HashMap weightsMap,
                               java.util.HashMap normWeights,
                               java.util.HashMap probabilities,
                               java.lang.String type,
                               int numNeighbors,
                               int hops)
                        throws java.lang.Exception
Iterative method for PRNMultiHop, uses a bipartite graph. All categories of an instance are considered (binary classification)

Parameters:
prncl - - Classifier in the inner loop
test - - Testset
id2ClassProb - - initialization of test instances
graph - - graph (jung)
maxiterations - - the maximal number of iterations to be performed
minWeight - - the minimal weight which an edge has to have in order to be considered in PRNMultiHop
weightsMap - - the weights of the edges in the graph
normWeights - - normalized weights
probabilities - - resulting probabilities for each class and test instance
type - - indicates the type of problem (heterogenous or homogenous)
numNeighbors - - determines the number of neighbors for which additional indirect neighbors will be considered using PRNMultiHop
hops - - maximal number of hops to be performed
Throws:
java.lang.Exception

relaxationLabelingMultiLabel

public void relaxationLabelingMultiLabel(relational.classifier.ProbRNMultiLabel2Hop prncl,
                                         weka.core.Instances test,
                                         java.util.HashMap id2ClassProb,
                                         edu.uci.ics.jung.graph.impl.SparseGraph graph,
                                         int label,
                                         java.util.HashMap expert,
                                         java.util.HashMap result,
                                         double th,
                                         int maxIterations,
                                         java.util.HashMap probabilities,
                                         int numNeighbors,
                                         java.util.HashMap weights,
                                         java.lang.String type)
                                  throws java.lang.Exception
Iterative method for PRN2Hop. All categories of an instance are considered (binary classification)

Parameters:
prncl -
test -
id2ClassProb -
graph -
label -
expert -
result -
th -
maxIterations -
probabilities -
numNeighbors -
weights -
Throws:
java.lang.Exception

relaxationLabelingMultiLabel

public void relaxationLabelingMultiLabel(ProbRNMultiHopMultiLabel prncl,
                                         weka.core.Instances test,
                                         java.util.HashMap id2ClassProb,
                                         edu.uci.ics.jung.graph.impl.SparseGraph graph,
                                         int label,
                                         java.util.HashMap expert,
                                         java.util.HashMap result,
                                         double th,
                                         double minWeight,
                                         int maxIterations,
                                         java.util.HashMap probabilities,
                                         int numNeighbors,
                                         java.util.HashMap weights,
                                         java.util.HashMap normWeights,
                                         java.lang.String type,
                                         int hops)
                                  throws java.lang.Exception
Iterative method for PRNMultiHop All categories of an instance are considered (binary classification)

Parameters:
prncl - - Classifier in the inner loop
test - - Testset
id2ClassProb - - initialization of test instances
graph - - graph (jung)
label - - category to be considered
expert - - expert judgement (concerning the categories)
result - - results (categories)
th - - Threshold
maxIterations - - the maximal number of iterations to be performed
probabilities - - resulting probabilities for each class and test instance
numNeighbors - - determines the number of neighbors for which additional indirect neighbors will be considered using PRNMultiHop
weights - - the weights of the edges in the graph
normWeights - - normalized weights
type - - indicates the type of problem (heterogenous or homogenous)
hops - - maximal number of hops to be performed
Throws:
java.lang.Exception

relaxationLabelingMultiLabel

public void relaxationLabelingMultiLabel(ProbRNMultiLabel prncl,
                                         weka.core.Instances test,
                                         java.util.HashMap id2ClassProb,
                                         edu.uci.ics.jung.graph.impl.SparseGraph graph,
                                         int label,
                                         java.util.HashMap expert,
                                         java.util.HashMap result,
                                         java.util.HashMap weights,
                                         double th,
                                         int maxIterations,
                                         java.util.HashMap probabilities,
                                         java.lang.String type)
                                  throws java.lang.Exception
Iterative method for Probabilistic Relational Neighbor Classifier. All categories of an instance are considered (binary classification)

Parameters:
prncl - - Classifier in the inner loop
test - - Testset
id2ClassProb - - initialization of test instances
graph - - graph (jung)
label - - HashMap, containing the categories
expert - - expert judgement (concerning the categories)
result - - results (categories)
weights - - the weights of the edges in the graph
th - - Threshold
maxIterations - - the maximal number of iterations to be performed
probabilities - -resulting probabilities for each class and test instance
type - -
Throws:
java.lang.Exception