Ph.D. student Steven Holtzen and his advisors Profs. Guy Van den Broeck and Todd Millstein won an ACM SIGPLAN Distinguished Paper Award for their OOPSLA 2020 paper “Scaling Exact Inference for Discrete Probabilistic Programs” (http://www.cs.ucla.edu/~todd/research/oopsla20.pdf).
Probabilistic programming languages (PPLs) are an expressive means of representing and reasoning for probabilistic models. However, in practice, PPLs struggle with the computational challenge of “probabilistic inference”, which is the task of computing the probability of an event in the model. This paper presents a new probabilistic programming language called Dice (http://dicelang.cs.ucla.edu). Dice features a new approach to exact probabilistic inference for discrete probabilistic programs, which are not well supported by existing approaches. Dice exploits program structure in order to factorize inference, enabling it to perform exact inference on probabilistic programs with hundreds of thousands of random variables.