{"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/go-gradient-for-expectation-based-objectives","title":"GO Gradient for Expectation-Based Objectives","arxiv_id":"1901.06020","date":"2019-01-17","proceeding":"ICLR 2019 5","authors":["Yulai Cong","Miaoyun Zhao","Ke Bai","Lawrence Carin"],"abstract":"Within many machine learning algorithms, a fundamental problem concerns\nefficient calculation of an unbiased gradient wrt parameters $\\gammav$ for\nexpectation-based objectives $\\Ebb_{q_{\\gammav} (\\yv)} [f(\\yv)]$. Most existing\nmethods either (i) suffer from high variance, seeking help from (often)\ncomplicated variance-reduction techniques; or (ii) they only apply to\nreparameterizable continuous random variables and employ a reparameterization\ntrick. To address these limitations, we propose a General and One-sample (GO)\ngradient that (i) applies to many distributions associated with\nnon-reparameterizable continuous or discrete random variables, and (ii) has the\nsame low-variance as the reparameterization trick. We find that the GO gradient\noften works well in practice based on only one Monte Carlo sample (although one\ncan of course use more samples if desired). Alongside the GO gradient, we\ndevelop a means of propagating the chain rule through distributions, yielding\nstatistical back-propagation, coupling neural networks to common random\nvariables.","url_abs":"http://arxiv.org/abs/1901.06020v1","url_pdf":"http://arxiv.org/pdf/1901.06020v1.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":"go-gradient-for-expectation-based-objectives","repo_url":"https://github.com/YulaiCong/GOgradient","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1901.06020","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}