Regarding this "Covariance matrix is not positive definite" about tribuo HOT 2 CLOSED

Covariance matrices need to be positive definite, and one of those necessary properties is symmetry so you need to at least make sure the matrix is symmetric. Assuming you're getting the error out of distribution2 which doesn't make the matrix symmetric unlike distribution1.

from tribuo.

Thanks a lot, now its work correctly.

package GeneratingData.org;

import org.tribuo.math.distributions.MultivariateNormalDistribution;
import tech.tablesaw.api.DoubleColumn;
import tech.tablesaw.api.Table;

import java.nio.file.Path;
import java.util.*;
import java.util.stream.Collectors;
import java.util.stream.IntStream;

/**

• This class is used to generate the dataset which is used for binary classification
  */
  public class GenerateDataSet {
  private final Path path;
  private final int numOfInstances;
  private final int numOfVariables;
  private final double P_Class1;
  private final Random random;

  /**

  • This is a default constructor for generating binary data
  • @param path The directory to store the generated data in
  • @param numOfInstances The number of samples per each class
  • @param numOfVariables The number of features for each class
  • @param priorC1 The prior for class C1
    */
    public GenerateDataSet(Path path, int numOfInstances, int numOfVariables, double priorC1) {
    this.path = path;
    this.numOfInstances = numOfInstances; // Number of samples for each class
    this.numOfVariables = numOfVariables; // Number of features for each class
    this.P_Class1 = priorC1;
    this.random = new Random();
    }

  /**

  • This method is called to generate the normally distributed data for two classes
    */
    public void generateDataSet() {
    // The number of samples in each class
    int numOfSamplesInC1 = (int) (numOfInstances * P_Class1);
    int numOfSamplesInC2 = (int) (numOfInstances * (1 - P_Class1));

    // The vector of mean values
    double[] meanVecC1 = new double[numOfVariables];
    for (int i = 0; i < meanVecC1.length; i++) {
    meanVecC1[i] = 1 + new Random().nextInt(4);
    }
    double[] meanVecC2 = new double[numOfVariables];
    for (int i = 0; i < meanVecC2.length; i++) {
    meanVecC2[i] = -1 * meanVecC1[i];
    }

    // The co_Variance matrices
    double[][] coVarMatC1 = generateSymmetricPositiveDefiniteMatrix(numOfVariables);
    Arrays.stream(coVarMatC1).forEach(i -> System.out.println(Arrays.toString(i)));

    MultivariateNormalDistribution distribution1 = new MultivariateNormalDistribution(meanVecC1, coVarMatC1, 12345);

    System.out.println();

    double[][] coVarMatC2 = generateSymmetricPositiveDefiniteMatrix(numOfVariables);
    Arrays.stream(coVarMatC2).forEach(i -> System.out.println(Arrays.toString(i)));

    MultivariateNormalDistribution distribution2 = new MultivariateNormalDistribution(meanVecC2, coVarMatC2, 12345);

    // The matrix to hold data for both classes
    double[][] columns = new double[numOfInstances][numOfVariables + 1]; // The entire dataset

    // For Class 0
    double[][] tabledDataC0 = new double[numOfSamplesInC1][numOfVariables];
    IntStream.range(0, tabledDataC0.length).parallel().forEach(i -> tabledDataC0[i] = distribution1.sampleVector().toArray());
    for (int i = 0; i < tabledDataC0.length; i++) {
    for (int j = 0; j < tabledDataC0[0].length; j++) {
    columns[i][j] = tabledDataC0[i][j];
    columns[i][numOfVariables] = 0;
    }
    }
    // For Class 1
    double[][] tabledDataC1 = new double[numOfSamplesInC2][numOfVariables];
    IntStream.range(0, tabledDataC1.length).parallel().forEach(i -> tabledDataC1[i] = distribution2.sampleVector().toArray());
    int row = 0;
    for (int i = tabledDataC1.length; i < (tabledDataC1.length * 2); i++) {
    for (int j = 0; j < tabledDataC1[0].length; j++) {
    columns[i][j] = tabledDataC1[row][j];
    columns[i][numOfVariables] = 1;
    }
    row++;
    }
    // Store columns as list of lists
    List<List> listMain = IntStream.range(0, columns[0].length).mapToObj(i -> Arrays.stream(columns).
    map(column -> column[i]).collect(Collectors.toList())).collect(Collectors.toList());

    // Creat list of objects from DoubleColumn to prepare data for TableSaw Library
    List doubleColumns = new ArrayList<>();
    for (int i = 0; i < listMain.size() - 1; i++) {
    doubleColumns.add(DoubleColumn.create("F" + i, listMain.get(i)));
    }
    doubleColumns.add(DoubleColumn.create("Label", listMain.get(listMain.size() - 1)));

    // Store this table as CSV file in the project directory
    Table table = Table.create("Data", new ArrayList<>(doubleColumns));
    table.write().csv(path.toString());
    }

  /**

  • This method generates a symmetric positive definite matrix
  • @param size The number of the features we want to generate
  • @return A symmetric positive definite matrix
    */
    private double[][] generateSymmetricPositiveDefiniteMatrix(int size) {
    SplittableRandom random = new SplittableRandom();
    // Generate symmetric matrix
    double[][] matrix = new double[size][size];
    for (int i = 0; i < size; i++) {
    for (int j = 0; j < size; j++) {
    matrix[i][j] = 1 + random.nextInt(10);
    }
    }
    for (int i = 0; i < size; i++) {
    for (int j = i + 1; j < size; j++) {
    matrix[j][i] = matrix[i][j];
    }
    }
    // Calculate the sum of the absolute of all non-diagonal values
    double sum = 0;
    for (int i = 0; i < size; i++) {
    for (int j = 1; j < size; j++) {
    if (i != j) {
    sum += Math.abs(matrix[i][j]);
    }
    }
    }
    // Make the matrix positive definite matrix
    for (int i = 0; i < size; i++) {
    matrix[i][i] = matrix[i][i] + 0.1 + sum;
    }
    return matrix;
    }
    }

from tribuo.