2013/03/27 by Judea Pearl, Pearl, Judea
Computer Science · #Advanced Graph Neural Networks #Artificial Intelligence (cs.AI) #Bayesian Modeling and Causal Inference #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences
paper · pdf · doi:10.48550/arxiv.1304.3102
openalex publication_date 2013/03/27 · openalex created_date 2022/10/04 · openalex updated_date 2026/07/28
This paper extends the applications of belief-networks to include the\nrevision of belief commitments, i.e., the categorical acceptance of a subset of\nhypotheses which, together, constitute the most satisfactory explanation of the\nevidence at hand. A coherent model of non-monotonic reasoning is established\nand distributed algorithms for belief revision are presented. We show that, in\nsingly connected networks, the most satisfactory explanation can be found in\nlinear time by a message-passing algorithm similar to the one used in belief\nupdating. In multiply-connected networks, the problem may be exponentially hard\nbut, if the network is sparse, topological considerations can be used to render\nthe interpretation task tractable. In general, finding the most probable\ncombination of hypotheses is no more complex than computing the degree of\nbelief for any individual hypothesis. Applications to medical diagnosis are\nillustrated.\n