国际会议演讲稿 .docx

《国际会议演讲稿 .docx》由会员分享，可在线阅读，更多相关《国际会议演讲稿 .docx（14页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、国际会议演讲稿自我介绍Thankyou,Mr./Ms.Chair./professorMynameissangqian.Iamveryhonoredtobeheretodooralpresentation.IamaMasterstudentfromHohaiUniversityandIamcurrentlydoingsomeresearchonphysicallayersecurity.Today,Iwouldliketosharewithyousomeofmyresearchonrelayselectionincooperativecommunication.(external/ek?st?

2、rn?l;?k?st?rn?l/)内容安排：Mypresentationincludesthesefiveparts.First,somebackgroundinformationaboutthisresearch;Second,systemmodelwehavedone;Third,NN-basedrelayselectionschemewehaveproposedForth,SimulationandresultsanalysisAndlast,someconclusionswehavegotP4Partone,introductionFirstly,Iwouldliketogiveyou

3、abitofbackground.Differingfromthetraditionalcryptographictechniquesbasedonsecretkeys,wecanmakeuseofwirelesschannelcharacteristicstoenhancephysicallayersecurity.Cooperativecommunicationhasbeenwidelyrecognizedasaneffectivewaytocombatwirelessfadingandprovidediversitygainwhichisoneoftheresearchhotspots.

4、Machinelearningasanemergingtechnologyhasbeenwidelyappliedinimageprocessing,cancerprediction,stockanalysisandotherfields.Sowhynottryitinwirelesscommunication?P5:Next,IwanttotalkalittlebitaboutpresentstudyRecentstudiesondeeplearningforwirelesscommunicationsystemshaveproposedalternativeapproachestoenha

5、ncecertainpartsoftheconventionalcommunicationsystemsuchasmodulationrecognition、channelencodinganddecoding、channelestimationanddetectionandanautoencoderwhichcanreplacethetotalsystemwithanovelarchitecture【modulationrecognition:AnNNarchitectureformodulationrecognitionthatconsistsofa4-layerNNandtwotwo-l

6、ayerNNs。channelencodinganddecoding：AplainDNNarchitectureforchanneldecodingtodecodekbitsmessagesfromNbitsnoisycodewords。channelestimationanddetection：Adense-Netforsymbol-to-symboldetectioncanadoptlongshort-termmemory(LSTM)todetectanestimatedsymbol.Autoencoder：theautoencodercanrepresenttheentirecommun

7、icationsystemandjointlyoptimizethetransmitterandreceiveroveranAWGNchannel.】P6Sowhydidweconductthisresearch?Well,wewanttoexploitthepotentialbenefitsofdeeplearninginenhancingphysicallayersecurityincooperative(?/k?p?r?t?v/?wirelesscommunicationandreducethefeedbackoverheadinlimitedspectrumresoucebyourou

8、rproposedscheme.P8Nowletmemoveontoparttwo-systemmodelHere,youcanseeafigurewhichisasystemmodel.Thisisthesource;thesearetherelaynodesandthisisthedestination,thisistheeavesdropperThewholeprocessofcooperativewirelesscommunicationcanbedividedintotwophases.Inthefirstphase,thesourcebroadcaststhesignaltothe

9、optimalrelaywhichguaranteesperfectsecurity.AsshowninFig1,hrepresentsafadingsri,coefficientofthechannelfromthesourcetotherelaynode(R.)iInthesecondphase,theoptimalrelayforwardsascaledversionofitsreceivedsignaltothedestinationinthepresenceoftheeavesdropper,wheretheoptimalrelayisconsideredtoadoptamplify

10、-and-forward(AF)relayscheme.Inthisfigure,hrepresentsafadingcoefficientofthechannelfromtherelayiRtotherdi,destinationgrepresentsafadingcoefficientofthechannelfromtherelayiRtothe,reieavesdropper.P9:Hereyoucanseesomefollowingexpressions.Iamnotgoingtowasteourprecioustimeonthelengthyderivation.Iwouldlike

11、toinviteyoutodirectlytakealookattheequationinitsfinalform.Thisistheoptimalindexoftheselectedrelaywiththeconventionalrelayselectionscheme.Amaongthisexpression(),=max,0sidieiCCC-representstheachievablesecrecyrateofsystemmodelwhenthe-thirelayisselected.P11Nowletmemovetopartthree-NN-basedRelaySelectionH

12、ereyoucanseeafigurewhichshowsconventional3-layerneuralnetwork.Itconsistsofinputlayer,hiddenlayer1,hidden(/h?dn/)layer2andoutputlayer.Neuralnetworkcanlearnfeaturesfromrawdataautomaticallyandadjustparameters/p?r?m?t?(r)z/asweightsandbiases.Incomplex(?/k?mpleks/)conditions(scenarios(/s?n?r?/),)Neuralne

13、tworkhaspromisingapplicationsinrelayselectionforseveralreasons.First,thedeepnetworkhassuperior(/su?p?r?/)learningabilitydespite(/d?spa?t/)thecomplexchannelconditions.Second,Neuralnetworkcanhandlelargedatasetsbecauseofdistributed(/d?str?bj?t?d/)andparallel(/p?r?lel/)computing(/k?mpju?t?/s,whichensure

14、computation(/k?mpj?te?(?)n/)speedandprocessingcapacity(?/k?p?s?t?/).Third,variouslibrariesorframeworks,suchasTensorFlow,Theano,andCaffegiveitwideapplicationsInthispaper,theproblemoftherelayselectionismodeledasamulti(/m?lt?/,ao)-classificationproblem.Weadoptsimpleneuralnetwork(NN)toselecttheoptimalre

15、laytoguaranteesperfectsecrecyperformanceofrelaycooperativecommunicationsystem.(enhancephysicallayersecurity)P12Beforetrainingtheclassificationmodel,weneedtomakesomepreparationfordeeplearningtoacquireatrainingsetandatestingset.First,weneedtoproducerealfeaturevectorforeachexampleaccordingtochannelstat

16、einformation;becausethechannelstateinformationmatricesiscomposedofcomplexnumbersbutfeaturevectorsaregenerallycomposedofrealnumbers.Soweneedtochangecomplexnumbersintorealnumberswithabsolute(/?bs?lu?t/)valueoperation.Moreover,inordertoimprovetheclassificationperformance(precision),itisnecessarytonorma

17、lizethefeaturevectors.Second,weneedtodesignkeyperformanceindicator(KPI).Inordertoeffectivelypreventtheeavesdropperfrominterceptinginformation,wechooseachievablesecrecyrateastheKPIofsystem.ThisKPIindicates(represents/shows)thedifferenceoftheachievableratefromthesourcetothedestinationandtheachievabler

18、atefromthesourcetotheeavesdropper.Third,wecanmakelabelsforexamplesaccordingtoKPI.theindexoftherelaywhichobtainsthemaximum(?/m?ks?m?m/)KPIisregardedastheclasslabeloftheexample.P13ClassificationmodelThispicture(isabout)showsthewholeprocessofbuildingclassificationmodel.Thewholeprocessofbuildingclassifi

19、cationcanbedividedintotwophases,namelytrainingphaseandtestingphase.Inthefirstphase,weneedtochooseInthesecondphase,wecanpredict(/pr?d?kt/)labelsofoptimalrelayaccordingtoinputdataandassessclassificationperformance.P15Nowletmemovetopartfour-SimulationandResults(?/r?z?lts/)AnalysisHere,youcanseeafigurew

20、hichshowstherelationshipbetweentheaverageandtheachievablesecrecyratewithdifferentnumbersofrelays.Inthisfigure,thebluelinerepresentstheconventionalrelayselectionschemeandtheredlinerepresentstheNN-basedscheme.Inthisfigure,asthenumbersofrelayandtheaveragetransmitpowerofthesourceincreases,theachievables

21、ecrecyrateincreasesaccordingly,.whichmeansincreasingthenumberofrelayscaneffectivelyimprovethesecretperformance.Theredlinearealmost(?/?lm?st/)closetotheblueline,【whchindicatesthatourproposedscheme(i.e.theNN-basedscheme)achievesalmostthesamesecrecyratesasthoseoftheconventionalschemeforallvaluesofN,】P1

22、6Thistableshowsthethenormalized(/?n?rm?la?zd/)meansquare(?/skwe?/)error(NMSES)valuesofdirrerentrelaynodes.ThevalueofNMSEmeanstheperformancedifferencebetweentheconventionalschemeandourproposedscheme.ThevaluesofNMSEarebelow(/b?l?/)negative(neg?t?v/)20(-20dB),whichvalidateseffectivenessofourproposedsch

23、emeagain.P17Now,letmemovetothelastpart-ConclusionOkay,nowwearegoingtotakealookatthelastpart-Conclusion.P18Wehavegotthefollowingconclusions.First,Incomplex(conditions)scenarios,Neuralnetworkhaspromisingapplicationsinrelayselectionforsuperiorlearningability,computationspeedandprocessingcapacity.Second

24、,Comparedwiththeconventionalrelayselectionscheme,ourproposedschemeachievesalmostthesamesecrecyperformance.Andlast,Ourproposedschemehasanadvantage(/?dv?nt?d?/)ofrelativelysmallfeedbackoverhead,indicatingthatproposedschemecanbeappliedtotheconditions(scenarioswherethefeedbackislimited.(Iftheconventiona

25、lschemeneedsfeedbackofNcomplexnumbers,NN-basedschemewillonlyneedfeedbackofNrealnumbers.Therefore,thefeedbackoverheadofourproposedschemeishalf(/h?f/?)ofthatoftheconventionalscheme,)Q&A1、计算复杂度ComputationalcomplexityThebiggestdrawbackisthehighlyselectioncomplexitieswithasmallnumberofrelaynodes.Ifnumber

26、ofrelaynodeisbig,itwillhaveaadvantage.Thisneedourfurtherresearch.Q:TheexperimentshowsthatsecrecyrateisalmostthesameastraditionalmethodandwhatisthepromotionofusingNNtorelayselection.(whatismeaningofintroducingNNtorelayselection)A:Thatourproposedscheme(i.e.theNN-basedscheme)achievesalmostthesameachiev

27、ablesecrecyrateasthatoftheconventionalschemeindicatesthatourproposedschemeiseffectiveanditcanselectoptimalrelaynodewhichobtainsmaximumachievablesecrecyrate.Onereason(thefirstreason)isthatAdoptingNNforrelayselectionisanovelidea.Anotherreasonisthatthespectrumresourceisrelativelimitedandourproposedsche

28、mehassmallfeedbackoverhead.Q:whatsthemeaningof“perfectsecrecyperformance?Whatsthemeaningof“Comparedtotheconventionalrelayselectionscheme?A:“perfectsecrecyperformancemeanstheachievablesecrecyrateisthebiggestonewhichcanenhancephysicallayersecurity.Infact,theconventionalrelayselectionschemeistheexhaust

29、ivesearch.Theindexofrelayselectionwiththisschemeisthebestone.Q:“ItisobviousthatthefeedbackoverheadofproposedschemeishalfofthatoftheconventionalschemeA:well,Letsmakeanassumption.IftheconventionalschemeneedsfeedbackofNcomplexnumbers,NN-basedschemewillonlyneedfeedbackofNrealnumbers.Therefore,thefeedbac

30、koverheadofourproposedschemeishalf(/h?f/?)ofthatoftheconventionalschemeQ:1)Thetrainingdatasetandthetestingdatasetaresocalledlegitimatechannelcomplexmatrixandthewiretapchannelcomplexmatrix.2)Howthesedatacanbeobtained/generated?3)WhyitisimportantandsignificantastheinputoftheNN-basedapproachandwhatistheoutput?Thewholeprocessisnotclear.