Abstract
This paper introduces a deterministic approximation algorithm with error guarantees for computing the probability of propositional formulas over discrete random variables. The algorithm is based on an incremental compilation of formulas into decision diagrams using three types of decompositions: Shannon expansion, independence partitioning, and product factorization. With each decomposition step, lower and upper bounds on the probability of the partially compiled formula can be quickly computed and checked against the allowed error.