Package org.apache.commons.rng.sampling

Class CompositeSamplers.SamplerBuilder<S>

  • Type Parameters:
    S - Type of sampler
    All Implemented Interfaces:
    CompositeSamplers.Builder<S>
    Enclosing class:
    CompositeSamplers
    private static final class CompositeSamplers.SamplerBuilder<S>
extends java.lang.Object
implements CompositeSamplers.Builder<S>
    Builds a composite sampler.

    A single builder can be used to create composites of different implementing classes which support different sampler interfaces. The type of sampler is generic. The individual samplers and their weights can be collected by the builder. The build method creates the discrete probability distribution from the weights. The final composite is created using a factory to create the class.

    • Method Detail

      • build

        public S build​(UniformRandomProvider rng)
        Builds the composite sampler. The rng is the source of randomness for selecting which sampler to use for each sample.

        Note: When the sampler is created the builder is reset to an empty state. This prevents building multiple composite samplers with the same samplers and their identical underlying source of randomness.

        If only one sampler has been added to the builder then the sampler is returned and the builder is reset.

        Specified by:
        build in interface CompositeSamplers.Builder<S>
        Parameters:
        rng - Generator of uniformly distributed random numbers.
        Returns:
        the sampler
        Throws:
        java.lang.IllegalStateException - if no samplers have been added to create a composite.
        See Also:
        CompositeSamplers.Builder.size()

      • reset

        private void reset()
        Reset the builder.

      • createDiscreteSampler

        private DiscreteSampler createDiscreteSampler​(UniformRandomProvider rng,
                                              double[] weights)
        Creates the discrete sampler of the enumerated probability distribution.

        If the specialisation is a shared state sampler the discrete sampler will be an instance of SharedStateDiscreteSampler.

        Parameters:
        rng - Generator of uniformly distributed random numbers.
        weights - Weight associated to each item.
        Returns:
        the sampler

      • uniform

        private static boolean uniform​(double[] values)
        Check if all the values are the same.

        Warning: This method assumes there are input values. If the length is zero an ArrayIndexOutOfBoundsException will be thrown.

        Parameters:
        values - the values
        Returns:
        true if all values are the same

      • sum

        private static double sum​(double[] values)
        Compute the sum of the values.
        Parameters:
        values - the values
        Returns:
        the sum

      • mean

        private static double mean​(double[] values)
        Compute the mean of the values. Uses a rolling algorithm to avoid overflow of a simple sum. This method can be used to compute the mean of observed counts for normalisation to a probability: 
         double[] values = ...;
 int n = values.length;
 double mean = mean(values);
 for (int i = 0; i < n; i++) {
     // Two step division avoids using the denominator (mean * n)
     values[i] = values[i] / mean / n;
 }

        Warning: This method assumes there are input values. If the length is zero an ArrayIndexOutOfBoundsException will be thrown.

        Parameters:
        values - the values
        Returns:
        the mean