[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_281":282},"routes/_layout",{"_11":210},"routes/entity.$type.$id",{"_11":12,"_13":14,"_15":16,"_17":18,"_19":20,"_21":22,"_23":24,"_25":26,"_27":28},"dehydratedState",{"_73":74,"_75":76},"type","paper","id","paper_W2141253686","section","technologies","entity",{"_15":16,"_29":30,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":38,"_39":-5,"_40":41,"_42":43,"_44":45,"_46":47,"_48":-5,"_49":50,"_51":30,"_52":53,"_54":55,"_56":57,"_58":59,"_60":61,"_62":-5,"_63":64,"_65":-5,"_66":67,"_68":30,"_69":-5},"canonical","https://www.hugmap.com/entity/paper/paper_W2141253686/technologies","siteBaseUrl","https://www.hugmap.com","sectionCount",3,"jsonLd","{\"@context\":\"https://schema.org\",\"@graph\":[{\"name\":\"Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure\",\"url\":\"https://www.hugmap.com/entity/paper/paper_W2141253686/technologies\",\"@type\":\"ScholarlyArticle\",\"headline\":\"Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure\",\"datePublished\":\"2002-01-01\",\"isPartOf\":{\"@type\":\"Periodical\",\"name\":\"Neural Computation\"}},{\"@type\":\"CollectionPage\",\"name\":\"Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure 相关技术\",\"description\":\"Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure 在 HugMap 中的相关技术，共 3 条，含结构化关联信息与详情链接。\",\"url\":\"https://www.hugmap.com/entity/paper/paper_W2141253686/technologies\",\"inLanguage\":\"zh-CN\"},{\"@type\":\"BreadcrumbList\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"首页\",\"item\":\"https://www.hugmap.com/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"论文\",\"item\":\"https://www.hugmap.com/search?type=paper\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure\",\"item\":\"https://www.hugmap.com/entity/paper/paper_W2141253686\"},{\"@type\":\"ListItem\",\"position\":4,\"name\":\"相关技术\",\"item\":\"https://www.hugmap.com/entity/paper/paper_W2141253686/technologies\"}]}]}","name","Adjusting the Outputs of a Classifier to New aPrioriProbabilities: A Simple Procedure","nameZh","aliases","description","descriptionZh","status","active","externalIds","{\"openalex\":\"W2141253686\",\"doi\":\"10.1162/089976602753284446\"}","properties","source","openalex","createdAt","2026-05-20T14:33:13.195326838","updatedAt","2026-05-20T14:33:14.182065683","doi","10.1162/089976602753284446","arxivId","openalexId","W2141253686","title","abstractText","It sometimes happens (for instance in case control studies) that a classifier is trained on a data set that does not reflect the true a priori probabilities of the target classes on real-world data. This may have a negative effect on the classification accuracy obtained on the real-world data set, especially when the classifier's decisions are based on the a posteriori probabilities of class membership. Indeed, in this case, the trained classifier provides estimates of the a posteriori probabilities that are not valid for this real-world data set (they rely on the a priori probabilities of the training set). Applying the classifier as is (without correcting its outputs with respect to these new conditions) on this new data set may thus be suboptimal. In this note, we present a simple iterative procedure for adjusting the outputs of the trained classifier with respect to these new a priori probabilities without having to refit the model, even when these probabilities are not known in advance. As a by-product, estimates of the new a priori probabilities are also obtained. This iterative algorithm is a straightforward instance of the expectation-maximization (EM) algorithm and is shown to maximize the likelihood of the new data. Thereafter, we discuss a statistical test that can be applied to decide if the a priori class probabilities have changed from the training set to the real-world data. The procedure is illustrated on different classification problems involving a multilayer neural network, and comparisons with a standard procedure for a priori probability estimation are provided. Our original method, based on the EM algorithm, is shown to be superior to the standard one for a priori probability estimation. Experimental results also indicate that the classifier with adjusted outputs always performs better than the original one in terms of classification accuracy, when the a priori probability conditions differ from the training set to the real-world data. The gain in classification accuracy can be significant.","publishedDate","2002-01-01","publishedYear",2002,"venue","Neural Computation","citationCount",328,"authors","keywords",[70,71,72],"pdfUrl","topVenue",false,"displayName","displayDescription","Fault Detection and Control Systems","Machine Learning and Algorithms","Neural Networks and Applications","mutations",[],"queries",[77],{"_78":79,"_80":81,"_82":83,"_84":85},"dehydratedAt",1780788131461,"state",{"_87":88,"_89":90,"_91":79,"_92":-5,"_93":94,"_95":94,"_96":94,"_97":-5,"_98":-5,"_99":67,"_35":100,"_101":102},"queryKey",[86,19,14,16],"queryHash","[\"portal\",\"entity\",\"paper\",\"paper_W2141253686\"]","portal","data",{"_19":20,"_103":104,"_105":106,"_107":108,"_109":110,"_18":111,"_62":112},"dataUpdateCount",1,"dataUpdatedAt","error","errorUpdateCount",0,"errorUpdatedAt","fetchFailureCount","fetchFailureReason","fetchMeta","isInvalidated","success","fetchStatus","idle","entityType","Paper","taxonomies",[168,169,170,171],"events",[],"articles",[],[141,142,143],[113,114,115],{"_15":135,"_29":136,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":137,"_39":-5,"_40":41,"_42":138,"_44":139,"_121":140,"_51":-5,"_122":-5,"_123":-5,"_124":-5,"_60":-5,"_125":-5,"_126":-5,"_127":-5,"_128":-5,"_129":-5,"_68":136,"_69":-5},{"_15":130,"_29":131,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":132,"_39":-5,"_40":41,"_42":133,"_44":134,"_121":-5,"_51":-5,"_122":-5,"_123":-5,"_124":-5,"_60":-5,"_125":-5,"_126":-5,"_127":-5,"_128":-5,"_129":-5,"_68":131,"_69":-5},{"_15":116,"_29":117,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":118,"_39":-5,"_40":41,"_42":119,"_44":120,"_121":-5,"_51":-5,"_122":-5,"_123":-5,"_124":-5,"_60":-5,"_125":-5,"_126":-5,"_127":-5,"_128":-5,"_129":-5,"_68":117,"_69":-5},"person_A5110457076","Marco Saerens","{\"openalex\":\"A5110457076\"}","2026-05-14T08:56:56.317185038","2026-05-23T03:44:15.255213254","orcid","affiliation","researchArea","paperCount","homepage","linkedin","twitter","github","hindex","person_A5041666358","Patrice Latinne","{\"openalex\":\"A5041666358\"}","2026-05-20T14:33:14.177295806","2026-05-20T14:33:14.214781356","person_A5052532082","Christine Decaestecker","{\"openalex\":\"A5052532082\",\"orcid\":\"0000-0002-5495-1931\"}","2026-05-20T14:33:15.226639965","2026-05-21T04:59:41.197684525","0000-0002-5495-1931",{"_15":163,"_29":71,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":164,"_39":-5,"_40":146,"_42":165,"_44":166,"_149":167,"_151":152,"_153":-5,"_154":155,"_156":-5,"_124":-5,"_157":-5,"_68":71,"_69":-5},{"_15":158,"_29":70,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":159,"_39":-5,"_40":146,"_42":160,"_44":161,"_149":162,"_151":152,"_153":-5,"_154":155,"_156":-5,"_124":-5,"_157":-5,"_68":70,"_69":-5},{"_15":144,"_29":72,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":145,"_39":-5,"_40":146,"_42":147,"_44":148,"_149":150,"_151":152,"_153":-5,"_154":155,"_156":-5,"_124":-5,"_157":-5,"_68":72,"_69":-5},"tech_oa_T10320","{\"taxonomy_id\":\"tax_rd_T10320\"}","openalex_topics","2026-05-17T09:00:01.255985187","2026-05-17T09:00:01.255987331","techType","architecture","domain","Physical Sciences","proposedYear","maturityLevel","growing","wikipediaUrl","patentCount","tech_oa_T10876","{\"taxonomy_id\":\"tax_rd_T10876\"}","2026-05-17T09:00:06.373016246","2026-05-17T09:00:06.373018471","concept","tech_oa_T12072","{\"taxonomy_id\":\"tax_rd_T12072\"}","2026-05-17T09:00:09.178457539","2026-05-17T09:00:09.178459468","algorithm",{"_15":200,"_29":201,"_31":-5,"_32":202,"_33":203,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":207,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":201,"_69":203},{"_15":195,"_29":72,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":196,"_39":-5,"_40":41,"_42":197,"_44":198,"_176":177,"_178":-5,"_179":26,"_180":181,"_182":199,"_184":94,"_185":186,"_187":94,"_188":-5,"_68":72,"_69":-5},{"_15":189,"_29":70,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":190,"_39":-5,"_40":41,"_42":191,"_44":192,"_176":177,"_178":-5,"_179":26,"_180":193,"_182":194,"_184":94,"_185":186,"_187":94,"_188":-5,"_68":70,"_69":-5},{"_15":172,"_29":71,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":173,"_39":-5,"_40":41,"_42":174,"_44":175,"_176":177,"_178":-5,"_179":26,"_180":181,"_182":183,"_184":94,"_185":186,"_187":94,"_188":-5,"_68":71,"_69":-5},"tax_rd_T12072","{\"openalex_topic\":\"T12072\"}","2026-05-14T00:28:24.28831512","2026-05-23T12:37:14.193563359","taxonomyType","RESEARCH_DOMAIN","code","level","parentId","tax_rd_1702","path","Physical Sciences/Computer Science/Artificial Intelligence/Machine Learning and Algorithms","sortOrder","isLeaf",true,"entityCount","children","tax_rd_T10876","{\"openalex_topic\":\"T10876\"}","2026-05-14T00:28:08.313109342","2026-05-23T12:21:13.230930673","tax_rd_2207","Physical Sciences/Engineering/Control and Systems Engineering/Fault Detection and Control Systems","tax_rd_T10320","{\"openalex_topic\":\"T10320\"}","2026-05-14T00:27:16.239175777","2026-05-23T12:36:45.268811201","Physical Sciences/Computer Science/Artificial Intelligence/Neural Networks and Applications","tax_ind_ai","人工智能",[208,209],"Artificial Intelligence","manual","INDUSTRY","tax_ind_root","科技行业/人工智能","AI","AI 技术",{"_73":211,"_75":212},[],[213],{"_78":214,"_80":215,"_82":216,"_84":217},1780788129374,{"_87":219,"_89":90,"_91":214,"_92":-5,"_93":94,"_95":94,"_96":94,"_97":-5,"_98":-5,"_99":67,"_35":100,"_101":102},[86,218],"[\"portal\",\"industries\"]","industries",[220,221,222,223,224,225,226,227,228,229,230,231],{"_15":200,"_29":201,"_31":-5,"_32":280,"_33":203,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":207,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":201,"_69":203},{"_15":276,"_29":277,"_31":-5,"_32":-5,"_33":278,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":279,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":277,"_69":278},{"_15":270,"_29":271,"_31":-5,"_32":272,"_33":273,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":274,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":271,"_69":273},{"_15":266,"_29":267,"_31":-5,"_32":-5,"_33":268,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":269,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":267,"_69":268},{"_15":262,"_29":263,"_31":-5,"_32":-5,"_33":264,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":265,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":263,"_69":264},{"_15":256,"_29":257,"_31":-5,"_32":258,"_33":259,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":260,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":257,"_69":259},{"_15":252,"_29":253,"_31":-5,"_32":-5,"_33":254,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":255,"_184":94,"_185":186,"_187":94,"_188":-5,"_68":253,"_69":254},{"_15":248,"_29":249,"_31":-5,"_32":-5,"_33":250,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":251,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":249,"_69":250},{"_15":244,"_29":245,"_31":-5,"_32":-5,"_33":246,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":247,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":245,"_69":246},{"_15":240,"_29":241,"_31":-5,"_32":-5,"_33":242,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":243,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":241,"_69":242},{"_15":236,"_29":237,"_31":-5,"_32":-5,"_33":238,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":239,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":237,"_69":238},{"_15":232,"_29":233,"_31":-5,"_32":-5,"_33":234,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":204,"_42":-5,"_44":-5,"_176":205,"_178":-5,"_179":90,"_180":206,"_182":235,"_184":94,"_185":67,"_187":94,"_188":-5,"_68":233,"_69":234},"tax_ind_smartphone","智能手机","Smartphone","科技行业/智能手机","tax_ind_robot","机器人","Robotics","科技行业/机器人","tax_ind_wearable","智能穿戴","Smart Wearables","科技行业/智能穿戴","tax_ind_energy","新能源与智能制造","New Energy & Smart Manufacturing","科技行业/新能源与智能制造","tax_ind_biotech","生物科技","Biotech","科技行业/生物科技","tax_ind_quantum","量子计算","Quantum Computing","科技行业/量子计算","tax_ind_blockchain","区块链",[261],"Blockchain","科技行业/区块链","Web3","tax_ind_enterprise","企业软件","Enterprise Software","科技行业/企业软件","tax_ind_security","网络安全","Cybersecurity","科技行业/网络安全","tax_ind_semi","半导体",[275],"Semiconductor","科技行业/半导体","芯片","tax_ind_cloud","云计算","Cloud Computing","科技行业/云计算",[208,209],"env",{"_283":24},"SITE_BASE_URL"]

Adjusting the Outputs of a Classifier to New a<i>Priori</i>Probabilities: A Simple Procedure 论文

Adjusting the Outputs of a Classifier to New a<i>Priori</i>Probabilities: A Simple Procedure · 相关技术

相关技术