{"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/the-reparameterization-trick-for-acquisition","title":"The reparameterization trick for acquisition functions","arxiv_id":"1712.00424","date":"2017-12-01","proceeding":null,"authors":["James T. Wilson","Riccardo Moriconi","Frank Hutter","Marc Peter Deisenroth"],"abstract":"Bayesian optimization is a sample-efficient approach to solving global\noptimization problems. Along with a surrogate model, this approach relies on\ntheoretically motivated value heuristics (acquisition functions) to guide the\nsearch process. Maximizing acquisition functions yields the best performance;\nunfortunately, this ideal is difficult to achieve since optimizing acquisition\nfunctions per se is frequently non-trivial. This statement is especially true\nin the parallel setting, where acquisition functions are routinely non-convex,\nhigh-dimensional, and intractable. Here, we demonstrate how many popular\nacquisition functions can be formulated as Gaussian integrals amenable to the\nreparameterization trick and, ensuingly, gradient-based optimization. Further,\nwe use this reparameterized representation to derive an efficient Monte Carlo\nestimator for the upper confidence bound acquisition function in the context of\nparallel selection.","url_abs":"http://arxiv.org/abs/1712.00424v1","url_pdf":"http://arxiv.org/pdf/1712.00424v1.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":"the-reparameterization-trick-for-acquisition","repo_url":"https://github.com/svedel/greattunes","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"bayesian-optimization","task_name":"Bayesian Optimization"},{"task_slug":"global-optimization","task_name":"global-optimization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.00424","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}