Scaling Decision-Theoretic Probabilistic Programming Through Factorization

Explore a groundbreaking probabilistic programming language called dappl in this 18-minute conference talk from ACM SIGPLAN. Discover how dappl models decision-making and solves maximum expected utility problems with exact precision. Learn about the language's functional design, featuring first-class decision-making, rewards, and probabilistic uncertainty. Understand the innovative reasoning-via-compilation strategy that enables scalable MEU reasoning and provides a flexible programming environment for complex real-world decision-making tasks. Examine the reduction of dappl MEU computation to a branch-and-bound algorithm over compiled Boolean formulas, and its proof of correctness against denotational semantics. Gain insights into dappl's expressiveness, which matches that of established decision-theoretic probabilistic graphical models like influence diagrams.

[DRAGSTERS] Scaling Decision--Theoretic Probabilistic Programming Through Factorization