Module es.upm.fi.cig.multictbnc
Package es.upm.fi.cig.multictbnc.sampling

Class DataSampler

java.lang.Object
es.upm.fi.cig.multictbnc.sampling.DataSampler
public class DataSampler extends Object
Implements methods for the generation and writing of datasets sampled from Multi-CTBNCs.

  • Constructor Summary

    Constructors
    Constructor
    Description
    DataSampler()
     

  • Method Summary

    Modifier and Type
    Method
    Description
    static void
    generateDataset(MultiCTBNC<CPTNode,CIMNode> multiCTBNC, int numSequences, double durationSequences, boolean isNoiseAdded, String destinationPath)
    Sample a dataset from the provided model.
    static void
    generateDataset(MultiCTBNC<CPTNode,CIMNode> multiCTBNC, int numSequences, double durationSequences, double percentageNoisyStates, double stdDeviationGaussianNoiseWaitingTime, String destinationPath)
    Sample a dataset from the provided model.
    protected static MultiCTBNC<CPTNode,CIMNode>
    generateModel(int numFeatureVariables, int numClassVariables, int cardinalityFeatureVariables, int cardinalityClassVariables, double probabilityEdgeClassSubgraph, double probabilityEdgeBridgeSubgraph, double probabilityEdgeFeatureSubgraph, int minIntensity, int maxIntensity, int maxNumParentsFeature, boolean differentStructurePerDataset, boolean forceExtremeProb, boolean[][] adjMatrix)
    Generates a Multi-CTBNC that can be used to sample data.
    static void
    generateRandomCIM(CIMNode node, double minIntensity, double maxIntensity)
    Generate a uniformly distributed random conditional intensity matrix for a node of a continuous-time Bayesian network.
    static void
    generateRandomCIMs(CTBN<CIMNode> ctbn, double minIntensity, double maxIntensity)
    Generate uniformly distributed random conditional intensity matrices for a continuous-time Bayesian network.
    static void
    generateRandomCPT(CPTNode node, boolean forceExtremeProb)
    Generate an uniformly distributed random conditional probability table for a Bayesian network node.
    static void
    generateRandomCPTs(BN<CPTNode> bn, boolean forceExtremeProb)
    Generate uniformly distributed random conditional probability tables for a Bayesian network.

    Methods inherited from class java.lang.Object

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

  • Constructor Details

    • DataSampler

      public DataSampler()

  • Method Details

    • generateDataset

      public static void generateDataset(MultiCTBNC<CPTNode,CIMNode> multiCTBNC, int numSequences, double durationSequences, boolean isNoiseAdded, String destinationPath)
      Sample a dataset from the provided model.
      Parameters:
      multiCTBNC - model from which datasets are sampled
      numSequences - number of sequences of the dataset
      durationSequences - duration of the sequences
      isNoiseAdded - True if noise is added to the datasets, False otherwise
      destinationPath - path where the dataset is saved

    • generateDataset

      public static void generateDataset(MultiCTBNC<CPTNode,CIMNode> multiCTBNC, int numSequences, double durationSequences, double percentageNoisyStates, double stdDeviationGaussianNoiseWaitingTime, String destinationPath)
      Sample a dataset from the provided model.
      Parameters:
      multiCTBNC - model from which datasets are sampled
      numSequences - number of sequences of the dataset
      durationSequences - duration of the sequences
      percentageNoisyStates - percentage of class variables' states and state transitions of feature variables which are randomly sampled.
      stdDeviationGaussianNoiseWaitingTime - standard deviation of the Gaussian distribution used to sample noise to be added to the waiting times of feature variables in a certain state
      destinationPath - path where the dataset is saved

    • generateRandomCIM

      public static void generateRandomCIM(CIMNode node, double minIntensity, double maxIntensity)
      Generate a uniformly distributed random conditional intensity matrix for a node of a continuous-time Bayesian network.
      Parameters:
      node - node of a continuous-time Bayesian network
      minIntensity - minimum value of the intensities
      maxIntensity - maximum value of the intensities

    • generateModel

      protected static MultiCTBNC<CPTNode,CIMNode> generateModel(int numFeatureVariables, int numClassVariables, int cardinalityFeatureVariables, int cardinalityClassVariables, double probabilityEdgeClassSubgraph, double probabilityEdgeBridgeSubgraph, double probabilityEdgeFeatureSubgraph, int minIntensity, int maxIntensity, int maxNumParentsFeature, boolean differentStructurePerDataset, boolean forceExtremeProb, boolean[][] adjMatrix)
      Generates a Multi-CTBNC that can be used to sample data.
      Parameters:
      numFeatureVariables - number of feature variables
      numClassVariables - number of class variables
      cardinalityFeatureVariables - cardinalities of the feature variables
      cardinalityClassVariables - cardinalities of the class variables
      probabilityEdgeClassSubgraph - probability of adding an edge in the class subgraph
      probabilityEdgeBridgeSubgraph - probability of adding an edge in the bridge subgraph
      probabilityEdgeFeatureSubgraph - probability of adding an edge in the feature subgraph
      minIntensity - minimum intensity
      maxIntensity - maximum intensity
      maxNumParentsFeature - maximum number of feature variables that can be parents of another feature variable
      differentStructurePerDataset - true to used the structure defined for a previous model, false otherwise.
      forceExtremeProb - true to force the probabilities of the CPTs to be extreme (0 to 0.3 or 0.7 to 1), false otherwise
      adjMatrix - adjacency matrix used to define the structure of the model (null to define the structure randomly)
      Returns:
      Multi-CTBNC generated

    • generateRandomCPTs

      public static void generateRandomCPTs(BN<CPTNode> bn, boolean forceExtremeProb)
      Generate uniformly distributed random conditional probability tables for a Bayesian network.
      Parameters:
      bn - a Bayesian network
      forceExtremeProb - true to force the probabilities to be extreme (0 to 0.3 or 0.7 to 1) if the size of the sample space of the class variables is 2, false otherwise

    • generateRandomCIMs

      public static void generateRandomCIMs(CTBN<CIMNode> ctbn, double minIntensity, double maxIntensity)
      Generate uniformly distributed random conditional intensity matrices for a continuous-time Bayesian network.
      Parameters:
      ctbn - continuous-time Bayesian network
      minIntensity - minimum value of the intensities
      maxIntensity - maximum value of the intensities

    • generateRandomCPT

      public static void generateRandomCPT(CPTNode node, boolean forceExtremeProb)
      Generate an uniformly distributed random conditional probability table for a Bayesian network node.
      Parameters:
      node - node
      forceExtremeProb - true to force the probabilities to be extreme (0 to 0.3 or 0.7 to 1) if the size of the sample space of the class variables is 2, false otherwise