{"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/optimizing-regularized-cholesky-score-for","title":"Optimizing regularized Cholesky score for order-based learning of Bayesian networks","arxiv_id":"1904.12360","date":"2019-04-28","proceeding":null,"authors":["Qiaoling Ye","Arash A. Amini","Qing Zhou"],"abstract":"Bayesian networks are a class of popular graphical models that encode causal\nand conditional independence relations among variables by directed acyclic\ngraphs (DAGs). We propose a novel structure learning method, annealing on\nregularized Cholesky score (ARCS), to search over topological sorts, or\npermutations of nodes, for a high-scoring Bayesian network. Our scoring\nfunction is derived from regularizing Gaussian DAG likelihood, and its\noptimization gives an alternative formulation of the sparse Cholesky\nfactorization problem from a statistical viewpoint, which is of independent\ninterest. We combine global simulated annealing over permutations with a fast\nproximal gradient algorithm, operating on triangular matrices of edge\ncoefficients, to compute the score of any permutation. Combined, the two\napproaches allow us to quickly and effectively search over the space of DAGs\nwithout the need to verify the acyclicity constraint or to enumerate possible\nparent sets given a candidate topological sort. The annealing aspect of the\noptimization is able to consistently improve the accuracy of DAGs learned by\nlocal search algorithms. In addition, we develop several techniques to\nfacilitate the structure learning, including pre-annealing data-driven tuning\nparameter selection and post-annealing constraint-based structure refinement.\nThrough extensive numerical comparisons, we show that ARCS achieves substantial\nimprovements over existing methods, demonstrating its great potential to learn\nBayesian networks from both observational and experimental data.","url_abs":"http://arxiv.org/abs/1904.12360v1","url_pdf":"http://arxiv.org/pdf/1904.12360v1.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":"optimizing-regularized-cholesky-score-for","repo_url":"https://github.com/yeqiaoling/ARCS-BN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.12360","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}