2016/10/18 by Giso H. Dal, Dal, Giso H., Peter Lucas +1
Computer Science · Decision Sciences · #Advanced Database Systems and Queries #Artificial Intelligence (cs.AI) #Bayesian Modeling and Causal Inference #Data Quality and Management #FOS: Computer and information sciences #Logic in Computer Science (cs.LO) #Logic, Reasoning, and Knowledge #Machine Learning and Algorithms
paper · pdf · doi:10.48550/arxiv.1610.05551
openalex publication_date 2016/10/18 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Recent work on weighted model counting has been very successfully applied to\nthe problem of probabilistic inference in Bayesian networks. The probability\ndistribution is encoded into a Boolean normal form and compiled to a target\nlanguage, in order to represent local structure expressed among conditional\nprobabilities more efficiently. We show that further improvements are possible,\nby exploiting the knowledge that is lost during the encoding phase and\nincorporating it into a compiler inspired by Satisfiability Modulo Theories.\nConstraints among variables are used as a background theory, which allows us to\noptimize the Shannon decomposition. We propose a new language, called Weighted\nPositive Binary Decision Diagrams, that reduces the cost of probabilistic\ninference by using this decomposition variant to induce an arithmetic circuit\nof reduced size.\n