[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-116978-en":3,"doc-seo-116978-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},116978,8796095360427,"Lucas Martin","https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d",8,"Research & Report","Sequential Domain Adaptation by Synthesizing Distributionally Robust Experts","Least squares models can underperform when trained with only a small set of labeled target-domain samples. Supervised domain adaptation seeks better predictive accuracy by leveraging additional labeled data from a source distribution that is close to the target. This work synthesizes a family of least squares estimator experts robust to moment conditions. Robustness is enforced via Kullback-Leibler or Wasserstein-type divergences, yielding efficient convex optimization solutions and sequential predictions via Bernstein online aggregation.","Sequential Domain Adaptation by Synthesizing Distributionally Robust Experts  \nBahar Taskesen 1 Man-Chung Yue 2 Jos Blanchet 3 Daniel Kuhn 1 Viet Anh Nguyen 3 4  \nAbstract  \nLeast squares estimators, when trained on a few target domain samples, may predict poorly. Supervised domain adaptation aims to improve the predictive accuracy by exploiting additional labeled training samples from a source distribution that is close to the target distribution. Given available data, we investigate novel strategies to synthesize a family of least squares estimator experts that are robust with regard to moment conditions. When these moment conditions are specified using Kullback-Leibler or Wasserstein-type divergences, we can find the robust estimatorsefficiently using convex optimization. We use the Bernstein online aggregation algorithm on the proposed family of robust experts to generate predictions for the sequential stream of target test samples. Numerical experiments on real data show that the robust strategies may outperform non-robust interpolations of the empirical least squares estimators.  \n1. Introduction  \nA natural approach to improving predictive performance in data-scarce tasks involves translating informative signals from a data-abundant source domain to the data-scarce target domain. This transfer of knowledge is commonly referred to as domain adaptation or transfer learning, and it is increasingly applied in a wide range of settings, see for example Wilson & Cook (2020); Chu & Wang (2018); Weiss et al. (2016) and Redko et al. (2019) .  \nWe consider the supervised domain adaptation setting with scarce labeled target data. The key challenge here is the  \n1Risk Analytics and Optimization Chair, ´Ecole Polytechnique Fdrale de Lausanne 2Department of Applied Mathematics, The Hong Kong Polytechnic University 3Department of Management Science and Engineering, Stanford University 4VinAIResearch, Vietnam. Correspondence to: Bahar Taskesen \u003Cbahar.taskesen@epfl.ch> .  \nProceedings of the 38 th International Conference on Machine Learning, PMLR 139, 2021 . Copyright 2021 by the author(s) .  \nabsence of meaningful data to tune any parameters. However, in many practically relevant applications, new data will arrive sequentially to enrich the information on the target domain. In this case, many online algorithms can be utilized to adaptively learn the best predictor on the target domain, which also guarantee optimal asymptotic regrets (Lattimore & Szepesvri, 2020) .  \nIn this paper, we take a pragmatic approach to resolve aspecific setup of the domain adaptation problem. We assume access to a scarce labelled target data, and the future target data arrives sequentially. For example, consider understanding the dynamics of ride-sharing platforms requires insights about the demand and supply from both sides of the market. These insights are signalled through the ride fares, which can be explained by characteristics such as the travel distances and the origin-destination pairs of the trips, the time of the day as well as the weather conditions. The capability to correctly predict ride fares directly translates into improved profit forecasts, and thus it vitally supports the growth of new-coming platforms. In a competitive market, a follower (e.g., Lyft) needs to target a slightly different market segment than the leader (e.g., Uber) who had entered earlier. Thus, the demand and supply characteristics for the follower may differ from those of the leader. Nevertheless, as both platforms provide on-demand transportation, it is reasonable to assume that their supply and demand dynamics are similar. The follower, who possesses limited data, can query demand on the leader's platform to collect data in order to leap forward in its predictive precision. Our approach to solve this problem is illustrated in Figure 1 and it consists of two components:  \n1. Expert Generation Module: This module generates a set of competitive experts E by fine-tuning the explana","cbCaihB5lih9Rggo","https://ap.wps.com/l/cbCaihB5lih9Rggo","pdf",774359,1,11,"English","en",105,"# Abstract\n## Introduction\n## Expert Generation Module\n## Expert Aggregation Module\n## Framework Architecture","[{\"question\":\"What problem does the paper address in domain adaptation?\",\"answer\":\"It addresses supervised domain adaptation when labeled target data is scarce and future target samples arrive sequentially, making it difficult to tune parameters from meaningful target information.\"},{\"question\":\"How are the distributionally robust experts constructed?\",\"answer\":\"The method synthesizes a family of least squares estimator experts using moment conditions specified by Kullback-Leibler or Wasserstein-type divergences, then identifies the robust estimators efficiently via convex optimization.\"},{\"question\":\"How are predictions made as target test samples arrive over time?\",\"answer\":\"Predictions are generated sequentially by applying the Bernstein online aggregation algorithm to the proposed family of robust experts, without re-adapting the experts during aggregation.\"}]","Sequential Domain Adaptation by Synthesizing Distributionally Robust Experts | PDF",1785672938,28,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"sequential-domain-adaptation-by-synthesizing-distributionally-robust-experts","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/sequential-domain-adaptation-by-synthesizing-distributionally-robust-experts/116978/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in domain adaptation?","Question",{"text":75,"@type":76},"It addresses supervised domain adaptation when labeled target data is scarce and future target samples arrive sequentially, making it difficult to tune parameters from meaningful target information.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How are the distributionally robust experts constructed?",{"text":80,"@type":76},"The method synthesizes a family of least squares estimator experts using moment conditions specified by Kullback-Leibler or Wasserstein-type divergences, then identifies the robust estimators efficiently via convex optimization.",{"name":82,"@type":73,"acceptedAnswer":83},"How are predictions made as target test samples arrive over time?",{"text":84,"@type":76},"Predictions are generated sequentially by applying the Bernstein online aggregation algorithm to the proposed family of robust experts, without re-adapting the experts during aggregation.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]