MichelangeRoll: Sculpting Rational Distributions Exactly and Efficiently
Abstract
Simulating an arbitrary discrete distribution using fair coin tosses incurs trade-offs between entropy complexity and space and time complexity. Shannon's theory suggests that tosses are necessary and sufficient, but does not guarantee exact distribution. Knuth and Yao showed that a decision tree consumes fewer than tosses for one exact sample. Draper and Saad's recent work addresses the space and time aspect, showing that tosses, memory, and operations are all it costs, where is the common denominator of the probability masses in and is the number of possible outcomes. In this paper, MichelangeRoll recycles leftover entropy to break the "" barrier. With memory, the entropy cost of generating a ongoing sequence of is reduced to per sample.
Cite
@article{arxiv.2507.00915,
title = {MichelangeRoll: Sculpting Rational Distributions Exactly and Efficiently},
author = {Jui-Hsiang Shao and Hsin-Po Wang},
journal= {arXiv preprint arXiv:2507.00915},
year = {2025}
}
Comments
14 pages, 7 figures, RANDOM says no so here