gaussian_mv#

Generates random numbers from multivariate normal distribution.

Description#

The class object is used in the generate function to provide random numbers with d-variate normal (Gaussian) distribution with mean (a) and variance-covariance matrix C, where aRd ; C is d x d symmetric positive-definite matrix.

The probability density function is given by:

\[f_{a, C} (x) = \frac{1}{\sqrt{\mathrm{det}(2\pi C)}} \exp (- \frac{1}{2} (x-a)^T C^{-1} (x-a))\]

where x``∈ ``Rd .

Matrix C can be represented as C = TTT, where T is a lower triangular matrix - Cholesky factor of C.

API#

Syntax#

namespace oneapi::mkl::rng {
  template<typename RealType = float, layout Layout = layout::packed,
           typename Method = gaussian_mv_method::by_default>
  class gaussian_mv {
  public:
    using method_type = Method;
    using result_type = RealType;
    static constexpr layout layout_type = Layout;

    explicit gaussian_mv(std::uint32_t dimen, std::vector<RealType> mean,
                         std::vector<RealType> matrix); // deprecated since oneMKL 2023.0
    explicit gaussian_mv(std::uint32_t dimen, sycl::span<RealType> mean,
                         sycl::span<RealType> matrix);
    explicit gaussian_mv(const param_type& pt);

    std::uint32_t dimen() const;
    std::vector<RealType> mean() const;
    std::vector<RealType> matrix() const;
    param_type param() const;
    void param(const param_type& pt);
  };
}

Devices supported: CPU and GPU.

Include Files#

  • oneapi/mkl/rng.hpp

Template Parameters#

Name

Description

typename RealType = float

Type of the produced values. The specific values are as follows:

float

double

layout Layout = layout::packed

Type of the matrix storage. The specific values are as follows:

layout::packed

layout::full

layout::diagonal

See brief descriptions of the methods in Distributions Template Parameter Method.

typename Method = oneapi::mkl::rng::gaussian_mv_method::by_default

Generation method. The specific values are as follows:

oneapi::mkl::rng::gaussian_mv_method::box_muller

oneapi::mkl::rng::gaussian_mv_method::box_muller2

oneapi::mkl::rng::gaussian_mv_method::icdf

See brief descriptions of the methods in Distributions Template Parameter Method.

Input Parameters#

Name

Type

Description

dimen

std::uint32_t

Dimension of output random vectors

mean

sycl::span<RealType> (float, double)

Mean span a of dimension d.

matrix

sycl::span<RealType> (float, double)

Variance-covariance matrix C.

Note

  • explicit gaussian_mv(std::uint32_t dimen, std::vector<RealType> mean, std::vector<RealType> matrix); is deprecated and will be removed in one of the next releases. Use explicit gaussian_mv(std::uint32_t dimen, sycl::span<RealType> mean, sycl::span<RealType> matrix); instead.

  • When passing a sycl::span that is constructed over a user’s memory to the constructor, users must manage the memory under sycl::span by themselves. They must not destroy the memory while data are processed.