Abstract
I will discuss how to reduce a continuous random variable to a small program using biased coin flips, and interesting theoretical questions about the number of required coins to do this. I might also discuss unrelated theoretical work asking which circuit languages that represent distributions support efficient marginal inference.