Notes for Intel® oneAPI Math Kernel Library Vector Statistics

Developer Reference

Contents

ChiSquare (VSL_RNG_METHOD_CHISQUARE_CHI2GAMMA)

Random number generator of the chi-square distribution with degrees of freedom
v
. You can generate any successful random number
x
 of the chi-square distribution as follows:
  1. If
    v
     = 1 and
    v
     = 3, a chi-square distributed number is generated as a sum of squares of
    v
     independent normal random numbers with parameter mean value
    a
     = 0 and standard deviation
    σ
     = 1.
  2. If
    v
     is even and 2 ≤
    v
     ≤ 16, a chi-square distributed random number is generated according to the formula:
    , where
    ui
     is a vector of successive random numbers uniformly distributed over the interval (0, 1).
  3. If
    v
     ≥ 17 or
    v
     is odd and 5 ≤
    v
     ≤ 15, the ChiSquare distribution is reduced to Gamma distribution with parameters shape
    α
     =
    v
    /2, offset
    a
     = 0, and scalefactor
    β
     = 2. Random numbers of the Gamma distribution are generated using method VSL_RNG_METHOD_GAMMA_GNORM.

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.