{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/ais-bn-an-adaptive-importance-sampling","title":"AIS-BN: An Adaptive Importance Sampling Algorithm for Evidential Reasoning in Large Bayesian Networks","arxiv_id":"1106.0253","date":"2011-06-01","proceeding":null,"authors":["J. Cheng","M. J. Druzdzel"],"abstract":"Stochastic sampling algorithms, while an attractive alternative to exact\nalgorithms in very large Bayesian network models, have been observed to perform\npoorly in evidential reasoning with extremely unlikely evidence. To address\nthis problem, we propose an adaptive importance sampling algorithm, AIS-BN,\nthat shows promising convergence rates even under extreme conditions and seems\nto outperform the existing sampling algorithms consistently. Three sources of\nthis performance improvement are (1) two heuristics for initialization of the\nimportance function that are based on the theoretical properties of importance\nsampling in finite-dimensional integrals and the structural advantages of\nBayesian networks, (2) a smooth learning method for the importance function,\nand (3) a dynamic weighting function for combining samples from different\nstages of the algorithm. We tested the performance of the AIS-BN algorithm\nalong with two state of the art general purpose sampling algorithms, likelihood\nweighting (Fung and Chang, 1989; Shachter and Peot, 1989) and self-importance\nsampling (Shachter and Peot, 1989). We used in our tests three large real\nBayesian network models available to the scientific community: the CPCS network\n(Pradhan et al., 1994), the PathFinder network (Heckerman, Horvitz, and\nNathwani, 1990), and the ANDES network (Conati, Gertner, VanLehn, and Druzdzel,\n1997), with evidence as unlikely as 10^-41. While the AIS-BN algorithm always\nperformed better than the other two algorithms, in the majority of the test\ncases it achieved orders of magnitude improvement in precision of the results.\nImprovement in speed given a desired precision is even more dramatic, although\nwe are unable to report numerical results here, as the other algorithms almost\nnever achieved the precision reached even by the first few iterations of the\nAIS-BN algorithm.","url_abs":"http://arxiv.org/abs/1106.0253v1","url_pdf":"http://arxiv.org/pdf/1106.0253v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"ais-bn-an-adaptive-importance-sampling","repo_url":"https://github.com/kgourgou/adaptive-importance-sampling-BN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":null,"task_name":"Pathfinder"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}