Improving the characteristics of pseudorandom number generators on short samples

Novikov Pavel Vladimirovich

The paper proposes an approach that improves the quality of pseudorandom numerical sequences on short samples. А discrete approximation of the continuous distribution function is made. Various rules for constructing histograms have been considered. An algorithm has been created that forms the best (according to the criterion of agreement) uniform distribution on short samples. The ways of applying the algorithm to generate other distributions are demonstrated.

Keywords:pseudorandom number generator, discrete approximation of the distribution function, agreement criterion, histogram, rule for choosing the number of intervals, inverse function method, frequency spectrum, the central limit theorem, the Box-Muller transformation.