Microprogramming of special instructions for sampling of random variates from any probability distribution is a means of increasing sampling speed. The diversity of sampling techniques is narrowed to one general algorithm: conditional bit sampling. Conditional bit sampling uses a high-speed uniform random number generator based on feedback shift registers to sample one bit at a time. The probability of a bit being a one in the j-th position of a binary expanded variate is stored in a table of conditional probabilities. A comparison with the pseudorandom number yields a one or zero. The table of conditional probabilities is generated once and passed through an instruction to the microprogram which performs the sampling. One user instruction is issued for each variate returned.
T. G. Lewis, "Microprogramming For Probability Distribution Sampling," Proceedings of the ACM Annual Conference, ACM 1972, vol. 1, pp. 582 - 589, Association for Computing Machinery (ACM), Aug 1972.
The definitive version is available at https://doi.org/10.1145/569971.569974
Electrical and Computer Engineering
Article - Conference proceedings
© 2023 Association for Computing Machinery (ACM), All rights reserved.
01 Aug 1972