电脑桌面
添加小米粒文库到电脑桌面
安装后可以在桌面快捷访问

maximum-flow-acceleration-by-traversing-tree-based-two-boundary-graph-contraction原版完整文件VIP免费

maximum-flow-acceleration-by-traversing-tree-based-two-boundary-graph-contraction原版完整文件_第1页
1/26
maximum-flow-acceleration-by-traversing-tree-based-two-boundary-graph-contraction原版完整文件_第2页
2/26
maximum-flow-acceleration-by-traversing-tree-based-two-boundary-graph-contraction原版完整文件_第3页
3/26
ExpertSystemsWithApplications225(2023)120151Maximumflowaccelerationbytraversingtreebasedtwo-boundarygraphcontractionWeiWeia,b,PengpengWanga,b,YaboDongc,∗aKeyLaboratoryofGrainInformationProcessingandControl(HenanUniversityofTechnology),MinistryofEducation,Zhengzhou450001,ChinabHenanProvincialKeyLaboratoryofGrainPhotoelectricDetectionandControl,HenanUniversityofTechnology,Zhengzhou450001,ChinacCollegeofcomputerscienceandtechnology,ZhejiangUniversity,Hangzhou310027,ChinaARTICLEINFOABSTRACTKeywords:Tree-cutmappingBoundarynodeMaximumflowGraphcontractionOverlayThemaximumflow(max-flow)problemissofundamentalthatthecorrespondingalgorithmshavemanyapplicationsinmanyscenarios.Acommonaccelerationstrategyistoreducegraphsizebycontractingsubgraphs,buttherearefewcandidatesubgraphswithnosizelimitthatcanguaranteeexactmax-flowsolutionsaftercontraction.Weaddanewsubgraphcontractionconditiontotheexistingconditionswithcorrespondingefficientlocatingalgorithm,andweproposeamax-flowaccelerationalgorithmusingtraversingtree-basedcontraction(TTCMax).TTCMaxcanworkefficientlyonlybyusingconnectivityinformation.Throughdepth-firsttraversing,itcancontracttwo-boundarygraphsofanysizeintoedges,andthenaccelerateexistingmax-flowalgorithmsusingcontractedgraphs.Large-scalerandomandreal-worldgraphexperimentstestitseffectandtheaccelerationratiocanbemorethan50timeswithlowgraphreductionoverhead,indicatingthattheproposedalgorithmcanbeusedasaneffectivecomplementtoexistingalgorithms.1.IntroductionAsafundamentalandwidelyinvestigatedproblem,themaximumflow(max-flow)problemanditsdualproblem,theminimumcut(min-cut)problemhavemanyapplicationsandareoftenusedassubroutinesinotheralgorithms(Sherman,2009).Manyadvanceshavebeenmadeinthedevelopmentofefficientalgorithmsforthisproblem(Goldberg&Rao,1998).Fastsolutionscanbefoundthroughalgorithm-orientedacceleration(Boykov&Kolmogorov,2004;Goldberg&Tarjan,2014;Verma&Dhruv,2012),orgraphdata-orientedacceleration(Gusfield,1990;Wei,Liu,&Zhang,2018;Zhang,Hua,Jiang,Zhang,&Chen,2011;Zhang,Xu,Hua,&Zhao,2012;Zhao,Su,Liu,&Zhang,2014;Zhao,Xu,Hua,&Zhang,2012).Asfordata-orientedalgorithms,thecoreideaistoreducethesizeoftheinput,suchasreducingthesizeofthegraphbycontractingsubgraphstonodes.Therearemanycontraction-basedalgorithmsthatexploitvarioussubgraphcontractionconditions.Unfortunately,mostofthesealgorithmscanonlyobtainapproximatedvaluewhenusingsubgraphcontractionconditionwithnosizelimit.Inthepaper,weproposealosslessalgorithmthatemploysasub-graphcontractionconditionwithnosizelimit,andatoyexampleisgiveninFig.1toclarifythedifferencebetweentheproposedmethodandexistingcontraction-basedmethods.Unlikeexistingnode-orientedcontractions,theproposedalgorithmcontractssubgraphsintoedges.Thecorepartisaroutinethatcanefficientlysearchforthetwo-boundarygraph(TBG),thatis,thesubgraphcontainingtwoboundarynodesthatseparateitfromotherpartsofthewholegraph.Experimentsvalidateitseffectivenessbyusinglarge-scalegraphscontainingupto107nodes.Theworkadvancesthestateoftheartbyintroducinganewlosslessandgeneralsize-freecontractionconditionwithahighlyefficientwaytolocatethecorrespondingsubgraphs.Specifically,thetheoreticalcon-tributionsinclude:(a)Weprovethatthecontractionoftwo-boundarygraphscanmaintaintheexactmax-flowresult,whichleadstothecontractionalgorithmthathasfewconstraints(e.g.,unlimitedsubgraphsize)andcanensureaccuratesolution...

1、当您付费下载文档后,您只拥有了使用权限,并不意味着购买了版权,文档只能用于自身使用,不得用于其他商业用途(如 [转卖]进行直接盈利或[编辑后售卖]进行间接盈利)。
2、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。
3、如文档内容存在违规,或者侵犯商业秘密、侵犯著作权等,请点击“违规举报”。

碎片内容

maximum-flow-acceleration-by-traversing-tree-based-two-boundary-graph-contraction原版完整文件

确认删除?
VIP
微信客服
  • 扫码咨询
会员Q群
  • 会员专属群点击这里加入QQ群
客服邮箱
回到顶部