欢迎来到天天文库
浏览记录
ID:34457691
大小:1.91 MB
页数:37页
时间:2019-03-06
《智能推荐算法》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、InformationFilteringonComplexNetworks:Prediction,RankingandRecommendationLinyuanLü吕琳媛E-mail:linyuan.lue@unifr.ch2012-5-31Dep.Phys.,Univ.Fribourg1OutlineòBackground&MotivationòWhatisLinkPredicting?òHowtoRankInfluentialNodes?òWhatisPersonalizedRecommendation?òRealApplicationsòOngoingwor
2、ks2012-5-31Dep.Phys.,Univ.Fribourg2BackgroundòEveryonecancreateinformationòMoreinformationareavailableConfusionMotivationInformationOverloadòOrganization&Reconstruction&Prediction2012-5-31Dep.Phys.,Univ.Fribourg3Prediction—completepartialinformationSamplenetworks:EvolvingnetworksPredi
3、ctionofexistedyetunknownPredictionoffuturelinks,links,e.g.foodwebs,protein-proteine.g.onlinefriendshipinteractionnetworks,metabolicnetworksnetworks.Ask:whichlinkwillexist?—whowillbefriend?TargetUserAsk:whowillberecommendedAsk:whoistheleadership?touser1asfriend?Ranking—understandtheRec
4、ommendation—significantpartbasedonknownguidetheinformationinformationgeneratingSubproblemofLP2012-5-31Dep.Phys.,Univ.Fribourg4Whyitissignificant?òPrediction——Howtocompleteinformation?òReduceexperimentalcosts(missinglinks)Discoveryoflinks/interactionsofbiologicalnetworkscostsmuch.Inste
5、adofblindlycheckingallpossiblelinks,topredictinadvanceandfocusonthemostlikelyexistinglinkscansharplyreducecostsifthepredictionsareaccurateenough.òOnlinefriendshiprecommendations(futurelinks)Recommendnewfriendstotheusersinwebsociety.òApplicationsinotherfieldsItcanbeappliedinsolvingthen
6、odeclassificationprobleminpartiallylabelednetworks,improvingrecommendationforsparsesystems,reconstrcuctingnoisynetworksòRanking——Howtoidentifytheimportantparts?òInformationsecurity:publicopinionssupervision,earlywarningandcontrol.òCommercialvalue:advertisementdelivery,newsspreading.òP
7、otentiallyenhancetheactivenessandadhesivecapacityofthesocialnetwork.òRecommendation——Howtofindwhattheuserslikeefficientlyandguidetheinformationgenerating?òE-Commercialapplications2012-5-31Dep.Phys.,Univ.Fribourg5CS&PhysicsPerspectivesMethodologiesWearetryingtoexplainwhyitworkslikethis
8、!2012
此文档下载收益归作者所有