[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-501586-105":59,"doc-detail-501586-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","good-learners-for-evil-teachers-supervised-machine-learning-algorithm","Good Learners for Evil Teachers - supervised machine learning algorithm","","Supervised machine learning often assumes training labels come from reliable, i.i.d. sources, but real data can be produced by heterogeneous teachers with varying competence and even deliberate malicious intent. This paper presents an SVM-based algorithm that reduces the influence of low-quality or malicious teachers without using prior teacher information or repeated labeling. A theoretical analysis establishes guarantees, and experiments demonstrate strong empirical performance, along with a second method with promising results but no formal proof.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/good-learners-for-evil-teachers-supervised-machine-learning-algorithm/501586/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/good-learners-for-evil-teachers-supervised-machine-learning-algorithm/501586.png","ImageObject",300,407,{"name":92,"@type":93},"dbqoiu","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-16","2026-10-03",true,{"@type":102,"interactionType":103,"userInteractionCount":29},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does the paper study in multi-teacher learning?","Question",{"text":112,"@type":113},"It studies learning with heterogeneous teachers that may be mediocre, incompetent, or even malicious, where labels can be low-quality or intentionally misleading.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does the proposed method handle malicious or low-quality teachers?",{"text":117,"@type":113},"It uses an algorithm built on the SVM framework that explicitly decreases the influence of low-quality and malicious teachers during learning, without prior knowledge of teachers.",{"name":119,"@type":110,"acceptedAnswer":120},"What data-labeling settings does the paper avoid compared with prior approaches?",{"text":121,"@type":113},"It avoids relying on prior information about teacher relationships and avoids repeated labeling where each example is labeled by multiple teachers.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},501586,1791432953,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":29,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":39,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":143,"read_time":46},3985747862462,"https://ap-avatar.wpscdn.com/davatar_3d24733baf745e90a7e4bdd5f77d97b2","Good Learners for Evil Teachers  \nOfer Dekel  \nMicrosoft Research, 1 Microsoft Way, Redmond, WA 98052, USA  \nOhad Shamir  \nOFERD @MICROSOFT. COM  \nOHADSH @CS . HUJI . AC. IL  \nThe Hebrew University, Jerusalem 91904, Israel  \nAbstract  \nWe consider a supervised machine learning scenario where labels are provided by a heterogeneous set of teachers, some of which are mediocre, incompetent, or perhaps even malicious. We present an algorithm, built on the SVM framework, that explicitly attempts to cope with low-quality and malicious teachers by decreasing their influence on the learning process. Our algorithm does not receive any prior information on the teachers, nor does it resort to repeated labeling (where each example is labeled by multiple teachers) . We provide a theoretical analysis of our algorithm and demonstrate its merits empirically. Finally, we present a second algorithm with promising empirical results but without a formal analysis.  \n1. Introduction  \nA supervised machine learning algorithm receives a training set of labeled examples and returns a hypothesis that attempts to accurately predict the labels of new unseen instances. When designing learning algorithms, we typically overlook the data collection process and make the simplistic assumption that the training set is sampled i.i.d. from some fixed distribution. However, real datasets are often non-homogeneous. In particular, training examples may be labeled by various different teachers. Labels provided by different teachers can be of different quality, due to varying degrees of expertise, competence, and dedication among teachers. Moreover, some teachers may deliberately try to manipulate or confuse our learning algorithms by providing incorrect labels.  \nFor example, imagine an algorithm that analyzes the logs generated by a search engine, with the intent of improving  \nAppearing in Proceedings of the 26 th International Conference on Machine Learning, Montreal, Canada, 2009 . Copyright 2009 by the author(s)/owner(s) .  \nfuture search engine results. A search-engine log records which links were clicked-on by each user. We think of each user as a distinct teacher, and of each click as a label. The click-patterns of most users are informative and helpful, while the click-patterns of other users merely introduce noise. Search-engine optimizers (SEOs) are individuals who try to reverse-engineer the algorithms used to construct a search engine and to manipulate these algorithms in their favor. If a SEO knows that our search engine promotes pages that received many historic clicks, he may masquerade as numerous users and simulate fictitious clicks on the links that he wants to promote. Identifying this form of click-spam and attenuating its effect is an essential step in the learning process.  \nAnother example of a multi-teacher scenario involves the use of crowdsourcing websites. Websites such as Galaxy Zoo (galaxy classification) and Stardust@home (interstellar dust particle detection) let members of the public volunteer their services by labeling astronomical images over the [Internet. Amazon.com](Internet. Amazon.com) ’s Mechanical Turk is an online system on which any individual can publish a crowdsourcing task and offer a payment for its completion. Typically, a dataset labeling task is broken up into multiple subtasks, and each subtask is completed by a different worker. Occasionally, workers are tempted to cheat by building automated systems, known as bots, that appear to solve the tasks but actually provide worthless labels. Although these bots do not directly try to harm our learning algorithm, they do try to trick us into believing that their labels are genuine. This intentional deceitful input can be as detrimental to our learning algorithm as deliberate malice. Weeding out the bots from among the human teachers is an important and difficult feat.  \nTo our knowledge, the existing literature on multi-teacher learning discusses two main approaches: using prior","cbCaikKXAx1MaRD5","https://ap.wps.com/l/cbCaikKXAx1MaRD5","pdf",170288,"English","# Introduction\n## Multi-teacher labeling and label quality\n## Motivating examples: search click-spam\n## Motivating examples: crowdsourcing and bots\n## Related work and limitations\n## Paper contributions","[{\"question\":\"What problem does the paper study in multi-teacher learning?\",\"answer\":\"It studies learning with heterogeneous teachers that may be mediocre, incompetent, or even malicious, where labels can be low-quality or intentionally misleading.\"},{\"question\":\"How does the proposed method handle malicious or low-quality teachers?\",\"answer\":\"It uses an algorithm built on the SVM framework that explicitly decreases the influence of low-quality and malicious teachers during learning, without prior knowledge of teachers.\"},{\"question\":\"What data-labeling settings does the paper avoid compared with prior approaches?\",\"answer\":\"It avoids relying on prior information about teacher relationships and avoids repeated labeling where each example is labeled by multiple teachers.\"}]","Good Learners for Evil Teachers - supervised machine learning algorithm | PDF",1791020976]