结合二部图投影与排序的协同过滤(3)

由结构相似得到的投影图可以用来描述原二部图所隐含的节点间的信息.但是这类相似计算方法是对称性的,认为σ(i,j)等于σ(j,i),即节点i,j视对方是同等重要的.其次,结构相似简单地视节点i,j的所有邻居节点对i,j的相似程度

Γi∩Γj,且ΓsµΓt,则显有相同的影响,然而,假设节点s,t∈

然邻居节点t应当比邻居结点s更能说明节点i,j相似.由于结构相似存在着以上种种缺点和不足,使其不能完全地反映原二部图所隐含的节点间的相似信息.为了更准确地挖掘原二部图的信息,寻找新的二部图投影方案将是未来工作的主要努力方向.

References:

[1]GedimiinasAdomavicus,AlexanderTuzhilin.Towardthenext

generationofrecommendersystems:asurveyofthestate2of2the2artandpossibleextensions[J].IEEETransactionsonKnowledgeandDataEngineering,June2005,17(6):7342749.

[2]LindenG,SmithB,recommendations:i2

tem2to2itemcollaborativefiltering[J].IEEEInternetComputing,2003,7(1):76280.

[3]MillerBN,AlbertI,LamSK,etal.MovieLensunplugged:ex2

perienceswithanoccasionallyconnectedrecommendersystem[C].ProceedingofInt′lConf.IntelligentUserInterfaces,Miami,USA:2003,2632266.

[4]ResnickP,IacovouN,SuchakM,etal.Grouplens:anopenar2

chitectureforcollaborativefilteringofnetnews[C].ProceedingoftheCSCWConference,ChapelHill,NC:1994,1752186.[5]KimJW,LeeBH,ShawMJ,etal.Applicationofdecision2tree

inductiontechniquestopersonalizedadvertisementsonInternetstorefronts[J].InternationalJournalofElectronicCommerce,2001,5(3):45262.

[6]BalabanovicM,ShohamY.Fab:content2based,collaborativerec2

ommendation[J].Comm.ACM,1997,40(3):66272.

[7]PazzaniM.Aframeworkforcollaborative,content2based,andde2

mographicfiltering[J].ArtificialIntelligenceRev.,Dec.1999,13:3932408.

[8]SouvikDebnath,NiloyGanguly,PabitraMitra.Featureweighting

incontentbasedrecommendationsystemusingsocialnetworkanal2ysis[C].ProceedingsofWWW,Beijing,China:April,2008,pages104121042.

[9]http://www.movielens.umn.edu.

[10]PavlovD,PennockD.Amaximumentropyapproachtocollabora2

tivefilteringindynamic,sparse,highdomains[C]rmationProcessingSystemsIPS′02),,2.

[11]reeseJD,panalysisofpredic2

sfor[C].ProceedingsofthethUinArtificialIntelligence,Madison:,252.[]D2Nowell,JonKleinberg.Thelink2predictionproblem

networks[J].JournaloftheAmericanSocietyforInfor2mationScienceandTechnology,2007,58:101921031.

[13]LeichtEA,PetterHolme,NewmanMEJ.Vertexsimilarityin

networks[J].PhysicalReviewE,2006,73:026120.

[14]BrinS,PageL.Theanatomyofalarge2scalehypertextualWeb

searchengine[J].ComputerNetworksandISDNSystems,1998,30(127):1072117.

[15]FoussF,PirotteA,SarensM.Random2walkcomputationofsimi2

laritiesbetweennodesofagraph,withapplicationtocollaborativerecommendation[J].IEEETransactionsonKnowledgeandDataEngineering,2007,19(3):3552369.

[16]LuhYen,MarcoSaerens,AminMantrach,etal.afamilyofdis2

similaritymeasuresbetweennodesgeneralizingbothshortest2pathandthecommute2timedistances[C].ProceedingsoftheACMSIGKDDInternationalConferenceonKnowledgeDiscoveryandDataMining(KDD),LasVegas,USA:2008,7852793.[17]MarcoGori,AugustoPucol.Arandom2walkbasedscoringalgo2

rithmwithapplicationtorecommendersystemsforlarge2scalee2commerce[C].ProceedingsofWEBKDD′06,Philadelphia,USA:2006,1272146.

[18]PanJY,YangHJ,FaloutsosC,etal.Automaticmultimedia

cross2modalcorrelationdiscovery[C].ProceedingsofACMIntl.

(ConferenceonKnowledgeDiscoveryandDataMiningKDD′

04),SeattleUSA:2004,6532658.[19]FoussF,PirotteA,RendersJM,etal.Anovelwayofcomputing

dissimilaritiesbetweennodesofagraph,withapplicationtocollab2orativefiltering[C].ProceedingsofIEEE/WIC/ACMInterna2tionalJointConferenceonWebIntelligence,Compiegne,France:2005,5502556.

[20]SiegelS,CastellanJ.Nonparametricstatisticsforthebehavioral

sciences[M].NewYork,USA:McGraw2HillCollege,1988.[21]KuoFF,ChiangMF,ShanMK,etal.Emotion2basedmusic

recommendationbyassociationdiscoveryfromfilmmusic[C].Proceedingsofthe13thAnnualACMInternationalConferenceonMultimedia.Singapore:2005,5072510.

[22]SarkarP,MooreA.Atractableapproachtofindingclosesttrun2

cated2commute2timeneighborsinlargegraphs[C].ProceedingsofUAI.Vancouver,BC,Canada:2007.