{"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/reducing-reparameterization-gradient-variance","title":"Reducing Reparameterization Gradient Variance","arxiv_id":"1705.07880","date":"2017-05-22","proceeding":"NeurIPS 2017 12","authors":["Andrew C. Miller","Nicholas J. Foti","Alexander D'Amour","Ryan P. Adams"],"abstract":"Optimization with noisy gradients has become ubiquitous in statistics and\nmachine learning. Reparameterization gradients, or gradient estimates computed\nvia the \"reparameterization trick,\" represent a class of noisy gradients often\nused in Monte Carlo variational inference (MCVI). However, when these gradient\nestimators are too noisy, the optimization procedure can be slow or fail to\nconverge. One way to reduce noise is to use more samples for the gradient\nestimate, but this can be computationally expensive. Instead, we view the noisy\ngradient as a random variable, and form an inexpensive approximation of the\ngenerating procedure for the gradient sample. This approximation has high\ncorrelation with the noisy gradient by construction, making it a useful control\nvariate for variance reduction. We demonstrate our approach on non-conjugate\nmulti-level hierarchical models and a Bayesian neural net where we observed\ngradient variance reductions of multiple orders of magnitude (20-2,000x).","url_abs":"http://arxiv.org/abs/1705.07880v1","url_pdf":"http://arxiv.org/pdf/1705.07880v1.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":"reducing-reparameterization-gradient-variance","repo_url":"https://github.com/andymiller/ReducedVarianceReparamGradients","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"variational-inference","task_name":"Variational Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.07880","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}