[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128619-en":3,"doc-seo-128619-105":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":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":11,"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},128619,962084925636,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d",8,"Research & Report","Comparison-Based Learning with Rank Nets - Adaptive search through comparisons","The work studies search in databases using comparisons only: a user is shown two objects and indicates which one is closer to an intended target. It focuses on adaptive strategies that exploit rank relationships and a prior distribution, without requiring access to actual pairwise distances. A new method, RankNetSearch, is proposed for targets whose distributions have bounded doubling constant, achieving query complexity close to the target entropy and thus near optimal. The analysis extends to noisy comparison oracles and includes empirical comparisons across multiple datasets.","Comparison-Based Learning with Rank Nets  \nAmin Karbasi [amin.karbasi@epfl.ch](amin.karbasi@epfl.ch)  \nEPFL, Lausanne, Switzerland  \nStratis Ioannidis [stratis.ioannidis@technicolor.com](stratis.ioannidis@technicolor.com)  \nTechnicolor, Palo Alto, USA  \nLaurent Massouli􀀓e [laurent.massoulie@technicolor.com](laurent.massoulie@technicolor.com)  \nTechnicolor, Paris, France  \nAbstract  \nWe consider the problem of search through comparisons, where a user is presented with two candidate objects and reveals which is closer to her intended target. We study adaptive strategies for 􀀌nding the target, that require knowledge of rank relationships but not actual distances between objects. We propose a new strategy based on rank nets, and show that for target distributions with abounded doubling constant, it 􀀌nds the target in a number of comparisons close to the entropy of the target distribution and, hence, of the optimum. We extend these results to the case of noisy oracles, and compare this strategy to prior art over multiple datasets.  \n1. Introduction  \nIn search through comparisons, a user locates a target object in a database as follows. At each step, the database presents two objects to the user, who then selects among the pair the object closest to the target that she has in mind. This process continues until, based on the user's answers, the database can uniquely identify the target she has in mind.  \nThis kind of interactive navigation, also known as exploratory search, has numerous real-life applications (Marchionini, 2006; Ruthven, 2008), such as navigation in a database of pictures of people photographed in an uncontrolled environment (Tschopp et al., 2011) . Automated methods may fail to extract meaningful features from such photos. Even if this were possible, in many practical cases, images with similar low-level descriptors may have very di􀀋erent semantic content,  \nAppearing in Proceedings of the 29 th International Conference on Machine Learning, Edinburgh, Scotland, UK, 2012 . Copyright 2012 by the author(s)/owner(s) .  \nand thus be perceived di􀀋erently by users (Smeulderset al. , 2000; Lew et al. , 2006) . On the other hand, a human can easily sort images of people w.r.t. their similarity to a given person, and her answers can be used to rank images in the database in terms of this similarity.  \nFormally, the human user's feedback can be modelled as a \\comparison oracle\" (Goyal et al. , 2008) . Assuming a database N endowed with a distance metric d, capturing the \\distance\" or \\dissimilarity\" between di􀀋erent objects, a comparison oracle answers questions of the kind: \\Between two objects x and y in N, which one is closest to t under the metric d?\".  \nIn this paper, we study algorithms for identifying an unknown target with as few queries to such an oracle as possible. Most importantly, the algorithms we consider do not rely on a priori knowledge of the distance between objects: they cannot access an embedding of N in a metric space, nor can they compute the distance between two objects. Decisions on which queries to submit to the oracle depend only on (a) ranking relationships between objects, which can indeed be obtained through a comparison oracle and (b) the prior distribution 􀀖 from which the target is sampled.  \nAs discussed in Section 3.3, content search through comparisons can be framed as an active learning problem. A well-known active learning algorithm is the Generalized Binary Search (GBS) or splitting algorithm (Dasgupta, 2005) . Using GBS to submit queries to the oracle locates the target in OPT 􀀁 􀀀Hmax (􀀖) + 1 􀀁 queries, where Hmax (􀀖) = maxx2supp(􀀖) log 􀀖~~ 1~~(x) and OP T is the number of queries submitted by an optimal algorithm. In practice, GBS performs very well in terms of query complexity, suggesting that this bound can be tightened. However, the computational complexity of GBS is 􀀂(n3 ) for n = jNj, which makes it intractable for most large databases.  \nRecently, Karbasi et al. (2011) proposed a","cbCailu2B9u0MM1Y","https://ap.wps.com/l/cbCailu2B9u0MM1Y","pdf",384920,2,1,"English","en",105,"# Introduction\n## Comparison oracle and exploratory search\n## Active learning connection to GBS\n## Contributions and paper organization\n# Related Work","[{\"question\":\"What is the comparison oracle used in this research?\",\"answer\":\"The comparison oracle answers which of two objects is closer to the target under an underlying distance metric, without revealing the actual distances.\"},{\"question\":\"How does RankNetSearch use prior information?\",\"answer\":\"RankNetSearch adapts queries based on rank relationships obtained from comparisons and on a prior distribution from which the target is sampled.\"},{\"question\":\"What performance guarantees are claimed for the new strategy?\",\"answer\":\"For target distributions with bounded doubling constant, RankNetSearch finds the target using a number of comparisons close to the entropy of the target distribution and near the optimum; results are also extended to noisy oracles.\"}]","Comparison-Based Learning with Rank Nets - Adaptive search through comparisons | PDF",1786002135,20,{"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},"comparison-based-learning-with-rank-nets-adaptive-search-through-comparisons","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/comparison-based-learning-with-rank-nets-adaptive-search-through-comparisons/128619/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-08-23","2026-08-06",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is the comparison oracle used in this research?","Question",{"text":75,"@type":76},"The comparison oracle answers which of two objects is closer to the target under an underlying distance metric, without revealing the actual distances.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does RankNetSearch use prior information?",{"text":80,"@type":76},"RankNetSearch adapts queries based on rank relationships obtained from comparisons and on a prior distribution from which the target is sampled.",{"name":82,"@type":73,"acceptedAnswer":83},"What performance guarantees are claimed for the new strategy?",{"text":84,"@type":76},"For target distributions with bounded doubling constant, RankNetSearch finds the target using a number of comparisons close to the entropy of the target distribution and near the optimum; results are also extended to noisy oracles.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"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":29,"slug":126},9,"Religion & Spirituality","religion-spirituality",{"id":29,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":29,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]