A method for the high-speed generation of random numbers with arbitrary distributions

✍ Scribed by T. Ishikawa; P.Y. Wang; K. Wakui; T. Yabe

• Publisher: Elsevier Science
• Year: 1992
• Tongue: English
• Weight: 666 KB
• Volume: 70
• Category: Article
• ISSN: 0010-4655
• DOI: 10.1016/0010-4655(92)90112-c

✦ Synopsis

A new method for the high-speed generation of random numbers with arbitrary distributions is proposed. The method stores random numbers in two memories, and then picks out two random numbers and assembles them, resulting in a new random number. When the method is. realized with Fortran, the method can generate random numbers with the normal distribution about eleven times faster than the method relying on the central limit theorem, and those with the Planck distribution about nine times faster than the acceptance-rejection method. The method is realized by hardware. The system generates 5 x 106 random numbers per second with any distribution. The statistical quality of these random numbers is similar to or better than that by the conventional methods.