Core Concepts
The average runtime of the BTPE algorithm for generating binomial random variates converges to a constant.
Abstract
In this technical report, Vincent A. Cicirello explores the average runtime behavior of the BTPE algorithm for generating binomial random variates using his open-source Java library ρµ. The report delves into Kachitvichyanukul and Schmeiser's formula for acceptance-rejection sampling iterations, analyzing its limit behavior as n approaches infinity. By instrumenting the Java implementation from the ρµ library, Cicirello experimentally validates the analysis. The study reveals that the average runtime of BTPE converges to a constant as n grows large, providing insights into the efficiency of binomial random variate generation algorithms.
Stats
E[I] = p4 * (n choose M) * r^M * (1 - r)^(n-M)
E[V] = 2p4 * (n choose M) * r^M * (1 - r)^(n-M)
Expected number of uniform variates required by BTPE: 3.801 for p = 10/n, 2.319 for p = 0.5
Quotes
"The average runtime of BTPE converges to a constant as n grows large."
"Experimental results confirm analytical predictions regarding uniform variates required by BTPE."