{"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/asynchronous-parallel-bayesian-optimisation","title":"Asynchronous Parallel Bayesian Optimisation via Thompson Sampling","arxiv_id":"1705.09236","date":"2017-05-25","proceeding":null,"authors":["Kirthevasan Kandasamy","Akshay Krishnamurthy","Jeff Schneider","Barnabas Poczos"],"abstract":"We design and analyse variations of the classical Thompson sampling (TS)\nprocedure for Bayesian optimisation (BO) in settings where function evaluations\nare expensive, but can be performed in parallel. Our theoretical analysis shows\nthat a direct application of the sequential Thompson sampling algorithm in\neither synchronous or asynchronous parallel settings yields a surprisingly\npowerful result: making $n$ evaluations distributed among $M$ workers is\nessentially equivalent to performing $n$ evaluations in sequence. Further, by\nmodeling the time taken to complete a function evaluation, we show that, under\na time constraint, asynchronously parallel TS achieves asymptotically lower\nregret than both the synchronous and sequential versions. These results are\ncomplemented by an experimental analysis, showing that asynchronous TS\noutperforms a suite of existing parallel BO algorithms in simulations and in a\nhyper-parameter tuning application in convolutional neural networks. In\naddition to these, the proposed procedure is conceptually and computationally\nmuch simpler than existing work for parallel BO.","url_abs":"http://arxiv.org/abs/1705.09236v1","url_pdf":"http://arxiv.org/pdf/1705.09236v1.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":"asynchronous-parallel-bayesian-optimisation","repo_url":"https://github.com/kirthevasank/gp-parallel-ts","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"bayesian-optimisation","task_name":"Bayesian Optimisation"},{"task_slug":"thompson-sampling","task_name":"Thompson Sampling"}],"methods":[{"method_slug":"ts","method_name":"TS"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.09236","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}