Class SobolSequenceGenerator

java.lang.Object
org.apache.commons.math3.random.SobolSequenceGenerator
All Implemented Interfaces:
RandomVectorGenerator

public class SobolSequenceGenerator extends Object implements RandomVectorGenerator
Implementation of a Sobol sequence.

A Sobol sequence is a low-discrepancy sequence with the property that for all values of N, its subsequence (x1, ... xN) has a low discrepancy. It can be used to generate pseudo-random points in a space S, which are equi-distributed.

The implementation already comes with support for up to 1000 dimensions with direction numbers calculated from Stephen Joe and Frances Kuo.

The generator supports two modes:

Since:
3.3
See Also:
  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    private static final int
    The number of bits to use.
    private int
    The current index in the sequence.
    private final int
    Space dimension.
    private final long[][]
    The direction vector for each component.
    private static final String
    Character set for file input.
    private static final int
    The maximum supported space dimension.
    private static final String
    The resource containing the direction numbers.
    private static final double
    The scaling factor.
    private final long[]
    The current state.
  • Constructor Summary

    Constructors
    Constructor
    Description
    SobolSequenceGenerator(int dimension)
    Construct a new Sobol sequence generator for the given space dimension.
    SobolSequenceGenerator(int dimension, InputStream is)
    Construct a new Sobol sequence generator for the given space dimension with direction vectors loaded from the given stream.
  • Method Summary

    Modifier and Type
    Method
    Description
    int
    Returns the index i of the next point in the Sobol sequence that will be returned by calling nextVector().
    private void
    initDirectionVector(int d, int a, int[] m)
    Calculate the direction numbers from the given polynomial.
    private int
    Load the direction vector for each dimension from the given stream.
    double[]
    Generate a random vector.
    double[]
    skipTo(int index)
    Skip to the i-th point in the Sobol sequence.

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
  • Field Details

    • BITS

      private static final int BITS
      The number of bits to use.
      See Also:
    • SCALE

      private static final double SCALE
      The scaling factor.
    • MAX_DIMENSION

      private static final int MAX_DIMENSION
      The maximum supported space dimension.
      See Also:
    • RESOURCE_NAME

      private static final String RESOURCE_NAME
      The resource containing the direction numbers.
      See Also:
    • FILE_CHARSET

      private static final String FILE_CHARSET
      Character set for file input.
      See Also:
    • dimension

      private final int dimension
      Space dimension.
    • count

      private int count
      The current index in the sequence.
    • direction

      private final long[][] direction
      The direction vector for each component.
    • x

      private final long[] x
      The current state.
  • Constructor Details

    • SobolSequenceGenerator

      public SobolSequenceGenerator(int dimension) throws OutOfRangeException
      Construct a new Sobol sequence generator for the given space dimension.
      Parameters:
      dimension - the space dimension
      Throws:
      OutOfRangeException - if the space dimension is outside the allowed range of [1, 1000]
    • SobolSequenceGenerator

      public SobolSequenceGenerator(int dimension, InputStream is) throws NotStrictlyPositiveException, MathParseException, IOException
      Construct a new Sobol sequence generator for the given space dimension with direction vectors loaded from the given stream.

      The expected format is identical to the files available from Stephen Joe and Frances Kuo. The first line will be ignored as it is assumed to contain only the column headers. The columns are:

      • d: the dimension
      • s: the degree of the primitive polynomial
      • a: the number representing the coefficients
      • m: the list of initial direction numbers
      Example:
       d       s       a       m_i
       2       1       0       1
       3       2       1       1 3
       

      The input stream must be an ASCII text containing one valid direction vector per line.

      Parameters:
      dimension - the space dimension
      is - the stream to read the direction vectors from
      Throws:
      NotStrictlyPositiveException - if the space dimension is < 1
      OutOfRangeException - if the space dimension is outside the range [1, max], where max refers to the maximum dimension found in the input stream
      MathParseException - if the content in the stream could not be parsed successfully
      IOException - if an error occurs while reading from the input stream
  • Method Details

    • initFromStream

      private int initFromStream(InputStream is) throws MathParseException, IOException
      Load the direction vector for each dimension from the given stream.

      The input stream must be an ASCII text containing one valid direction vector per line.

      Parameters:
      is - the input stream to read the direction vector from
      Returns:
      the last dimension that has been read from the input stream
      Throws:
      IOException - if the stream could not be read
      MathParseException - if the content could not be parsed successfully
    • initDirectionVector

      private void initDirectionVector(int d, int a, int[] m)
      Calculate the direction numbers from the given polynomial.
      Parameters:
      d - the dimension, zero-based
      a - the coefficients of the primitive polynomial
      m - the initial direction numbers
    • nextVector

      public double[] nextVector()
      Generate a random vector.
      Specified by:
      nextVector in interface RandomVectorGenerator
      Returns:
      a random vector as an array of double.
    • skipTo

      public double[] skipTo(int index) throws NotPositiveException
      Skip to the i-th point in the Sobol sequence.

      This operation can be performed in O(1).

      Parameters:
      index - the index in the sequence to skip to
      Returns:
      the i-th point in the Sobol sequence
      Throws:
      NotPositiveException - if index < 0
    • getNextIndex

      public int getNextIndex()
      Returns the index i of the next point in the Sobol sequence that will be returned by calling nextVector().
      Returns:
      the index of the next point